Skip to main content

Full text of "books culture"

See other formats

A Companion to 
Analytic Philosophy 


Blackwell Companions to Philosophy 

This outstanding student reference series offers a comprehensive and authoritative 
survey of philosophy as a whole. Written by today's leading philosophers, each volume 
provides lucid and engaging coverage of the key figures, terms, topics, and problems of 
the field. Taken together, the volumes provide the ideal basis for course use, represent- 
ing an unparalleled work of reference for students and specialists alike. 

Already published in the series 

1. The Blackwell Companion to Philosophy 
Edited by Nicholas Bunnin and Eric 

2. A Companion to Ethics 
Edited by Peter Singer 

3 . A Companion to Aesthetics 
Edited by David Cooper 

4. A Companion to Epistemology 

Edited by Jonathan Dancy and Ernest Sosa 

5. A Companion to Contemporary Political 

Edited by Robert E. Goodin and Philip Pettit 

6. A Companion to Philosophy of Mind 
Edited by Samuel Guttenplan 

7. A Companion to Metaphysics 
Edited by Jaegwon Kim and Ernest Sosa 

8. A Companion to Philosophy of Law and 
Legal Theory 

Edited by Dennis Patterson 

9. A Companion to Philosophy of Religion 
Edited by Philip L. Quinn and Charles 

10. A Companion to the Philosophy of 

Edited by Bob Hale and Crispin Wright 

11. A Companion to World Philosophies 
Edited by Eliot Deutsch and Ron Bontekoe 

12. A Companion to Continental Philosophy 
Edited by Simon Critchley and William 

13. A Companion to Feminist Philosophy 
Edited by Alison M. Jaggar and Iris 
Marion Young 

14. A Companion to Cognitive Science 
Edited by William Bechtel and George 

15. A Companion to Bioethics 

Edited by Helga Kuhse and Peter Singer 

16. A Companion to the Philosophers 
Edited by Robert L. Arrington 

17. A Companion to Business Ethics 
Edited by Robert E. Frederick 

18. A Companion to the Philosophy of 

Edited by W. H Newton-Smith 

19 . A Companion to Environmental 

Edited by Dale Jamieson 

20. A Companion to Analytic Philosophy 
Edited by A. P. Martinich and David Sosa 


A Companion to Genethics 

Edited by John Harris and Justine Burley 

A Companion to African-American 


Edited by Tommy Lott and John Pittman 

A Companion to African Philosophy 
Edited by Kwasi Wiredu 

A Companion to Ancient Philosophy 
Edited by Mary Louise Gill 

A Companion to Early Modern 


Edited by Steven Nadler 

A Companion to Philosophical Logic 
Edited by Dale Jacquette 

A Companion to Medieval Philosophy 
Edited by Jorge J. E. Gracia, Greg 
Reichberg, and Timothy Noone 


Companions to 


A Companion to 
Analytic Philosophy 

Edited by 





Copyright © Blackwell Publishers Ltd 2001 

First published 2001 

2468 10 97531 

Blackwell Publishers Inc. 

350 Main Street 

Maiden, Massachusetts 02148 


Blackwell Publishers Ltd 
108 Cowley Road 
Oxford OX4 1JF 


All rights reserved. Except for the quotation of short passages for the purposes of 
criticism and review, no part of this publication may be reproduced, stored in a retrieval 
system, or transmitted, in any form or by any means, electronic, mechanical, 
photocopying, recording, or otherwise, without the prior permission of the publisher. 

Except in the United States of America, this book is sold subject to the condition that it 
shall not, by way of trade or otherwise, be lent, resold, hired out, or otherwise circulated 
without the publisher's prior consent in any form of binding or cover other than that in 
which it is published and without a similar condition including this condition being 
imposed on the subsequent purchaser. 

Library of Congress Cataloging-in-Publkation Data 

A companion to analytic philosophy / edited by A. P. Martinich and David Sosa. 
p. cm. - (Blackwell companions to philosophy) 
Includes bibliographical references and index. 
ISBN 0-631-21415-1 (hb : alk. paper) 

1. Analysis (Philosophy) 2. Philosophy, Modern - 19th century. 3. Philosophy, 
Modern - 20th century. I. Martinich, Aloysius. II. Sosa, David, 1966- III. Series. 
B808.5 .C555 2001 


British Library Cataloguing in Publication Data 

A CIP catalogue record for this book is available from the British Library. 

Typeset in 10 on 12.5 pt Photina 

by Best-set Typesetter Ltd., Hong Kong 

Printed in Great Britain by MPG Books Ltd, Bodmin, Cornwall 

This book is printed on acid-free paper. 


List of Contributors viii 

Introduction 1 


1 Gottlob Frege (1848-1925) 6 


2 Bertrand Russell (1872-1970) 21 


3 G. E. Moore (1873-1958) 45 


4 CD. Broad (1887-1971) 57 


5 Ludwig Wittgenstein (1889-1951) 68 


6 Rudolf Carnap (1891-1970) 94 


7 Karl Popper (1892-1994) 110 


8 Gilbert Ryle (1900-1976) 117 


9 Alfred Tarski (1902-1983), Alonzo Church (1903-1995), 

and Kurt Godel (1906-1978) 124 


10 Frank P. Ramsey (1903-1930) 139 


11 Carl G. Hempel (1905-1997) 148 



12 Nelson Goodman (1906-1998) 160 


13 H. L. A. Hart (1907-1992) 169 


14 Charles Stevenson (1908-1979) 175 


15 W. V. Quine (1908-2000) 181 


16 A. J. Ayer (1910-1989) 205 


17 J. L.Austin (1911-1960) 218 


18 Norman Malcolm (1911-1990) 231 


19 Wilfrid Sellars (1912-1989) 239 


20 H. P. Grice (1913-1988) 254 


21 G. H. von Wright (1916- ) 274 


22 Roderick Chisholm (1916-1999) 281 


23 Donald Davidson (1917- ) 296 


24 G. E. M. Anscombe (1919-2001) 315 


25 R. M. Hare (1919- ) 326 


26 P. F. Strawson (1919- ) 334 


27 Philippa Foot (1920- ) 350 


28 Ruth Barcan Marcus (1921- ) 357 


29 John Rawls (1921- ) 361 



30 Thomas S.Kuhn (1922-1996) 371 


31 Michael Dummett (1925- ) 378 


32 Hilary Putnam (1926- ) 393 


33 David M. Armstong (1926- ) 413 


34 Noam Chomsky (1928- ) 419 


35 Richard Rorty (1931- ) 428 


36 John R. Searle (1932- ) 434 


37 Jerry Fodor (1935- ) 451 


38 Saul Kripke (1940- ) 466 


39 David Lewis (1941- ) 478 


Index 489 


C. Anthony Anderson 

Professor of Philosophy, University of California, Santa Barbara 

Brad Armendt 

Associate Professor of Philosophy, Arizona State University 

Thomas Baldwin 

Professor of Philosophy, University of York, England 

Max Cresswell 

Professor of Philosophy, Victoria University of Wellington, New Zealand 

Norman Daniels 

Professor of Philosophy and Goldthwaite Professor of Rhetoric, Tufts University 

James Dreier 

Associate Professor of Philosophy, Brown University 

Sir Michael Dummett 

Wykeham Professor of Logic Emeritus, University of Oxford 

Richard Foley 

Professor of Philosophy and Dean of the Faculty of Arts and Sciences, New York University 

Carl Ginet 

Professor Emeritus of Philosophy, Cornell University 

Richard Grandy 

Carolyn and Fred McManis Professor of Philosophy, Rice University 

P. M. S. Hacker 

Fellow of St. John's College, University of Oxford 

John Heil 

Professor of Philosophy, Davidson College 

Peter Hylton 

Professor of Philosophy, University of Illinois, Chicago 



Frank Jackson 

Professor of Philosophy in the Philosophy Program, Research School of Social Sciences, and 
Director of the Institute of Advanced Studies, Australian National University 

Philip Kitcher 

Professor of Philosophy, Columbia University 

Gavin Lawrence 

Associate Professor of Philosophy, University of California, Los Angeles 

Ernest LePore 

Professor of Philosophy and Director of the Center for Cognitive Science, Rutgers University 

Peter Ludlow 

Professor of Philosophy, State University of New York, Stony Brook 

A. P. Martinich 

Roy Allison Vaughan Centennial Professor of Philosophy, Professor of History, University of 
Texas, Austin 

Alexander Miller 

Senior Research Fellow, Cardiff University 

Anselm Winfried Miiller 

Professor of Philosophy, University of Trier, Germany 

Stephen Neale 

Professor of Philosophy, Rutgers University 

W. H. Newton-Smith 

Fairfax Fellow, Jowett Fellow, Jowett Lecturer and Tutor in Philosophy, Balliol College, 
University of Oxford 

Georges Rey 

Professor of Philosophy, University of Maryland, College Park 

Jay F. Rosenberg 

Taylor Grandy Professor of Philosophy, University of North Carolina, Chapel Hill 

Sahotra Sarkar 

Associate Professor of Philosophy, University of Texas, Austin 

Israel Scheffler 

Professor Emeritus of Philosophy and Professor Emeritus and Director, Philosophy of 
Education Research Center, Graduate School of Education, Harvard University 

John R. Searle 

Mills Professor of Philosophy, University of California, Berkeley 

Scott Shapiro 

Assistant Professor of Law, Benjamin Cardozo School of Law, Yeshiva University 

Walter Sinnott-Armstrong 

Professor of Philosophy, Dartmouth College 


P. F. Snowdon 

Lecturer in Philosophy and Fellow, Exeter College, University of Oxford 

David Sosa 

Associate Professor of Philosophy, University of Texas, Austin 

Ernest Sosa 

Romeo Elton Professor of Natural Theology, Brown University, and Professor of Philosophy, 
Rutgers University 

T. L. S. Sprigge 

Honorary Fellow, University of Edinburgh 

Robert Stalnaker 

Professor of Philosophy, Massachusetts Institute of Technology 

Frederick Stoutland 

Professor Emeritus of Philosophy, St. Olaf College, and Permanent Visiting Professor, 
University of Uppsala 

Avrum Stroll 

Research Professor, University of California, San Diego 

James van Cleve 

Professor of Philosophy, Brown University 

Michael Williams 

Professor of Philosophy, Johns Hopkins University 

Dean Zimmerman 

Associate Professor of Philosophy, Syracuse University 

Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 



Though analytic philosophy was practiced by Plato and reinvigorated in the modern 
era by Rene Descartes and Thomas Hobbes among others, we are concerned with it 
only in its twentieth-century forms. As such, it was revived in two centers, Germany 
and England. In Germany, Gottlob Frege was exploring the foundations of math- 
ematics and logic. His efforts introduced new standards of rigor that made their way 
into analytic philosophy generally, through the work of Bertrand Russell and Ludwig 
Wittgenstein. His discussions of the nature of language and reasoning have also 
become powerful tools in the hands of later philosophers. Among Frege's many books 
and articles, the Grundgesetze, Begriffsschrift, "On Sense and Reference" ("Uber Sinn und 
Bedeutung," 1892) and "Thoughts" ("Gedanken," 1918) stand out as especially 

During about the same period in England, G. E. Moore led the way in opposing the 
then-dominant philosophy of British idealism. While "The Nature of Judgment" is an 
early criticism of a point in F. H. Bradley's Logic, the locus classicus of British analytic 
philosophy is likely "The Refutation of Idealism" (1903), a criticism of the formula esse 
est percipi ("to be is to be perceived"). A crucial part of that argument is Moore's claim 
that the concept of the sensation of yellow contains two parts: the sensation that is 
unique to each person and the yellowness that can be perceived by many people. Even 
when idealists conceded that there was some kind of duality here, they insisted on a 
kind of inseparability. 

To use a general name for the kind of analytic philosophy practiced during the first 
half of the twentieth century, initially in Great Britain and German-speaking countries, 
and later in North America, Australia, and New Zealand, "conceptual analysis" aims 
at breaking down complex concepts into their simpler components. Successive analy- 
ses performed on complex concepts would yield simpler concepts. According to Moore, 
the process might lead ultimately to simple concepts, of which no further analysis could 
be given. The designation "conceptual" was supposed to distinguish the philosophical 
activity from various analyses applied to nonconceptual objects. Physics was famous in 
the twentieth century for breaking down atoms into protons, neutrons, and electrons, 
and these subatomic particles into an array of more exotic components. And analytic 
chemistry aims at determining chemical compositions. The analogy between philoso- 
phy and science inspired the name "logical atomism," a theory that flourished between 



1920 and 1930. Both Wittgenstein and Bertrand Russell maintained that there 
must be simple, unanalyzable objects at the fundamental level of reality. Wittgenstein 
thought that the simples existed independently of human experience, Russell that they 
existed only for as long as one's attention was fixed on them. 

Notwithstanding the analogy between scientific and philosophical analysis, most 
philosophers in the first half of the twentieth century maintained that philosophy was 
very different from science. In his Tractatus Logico-Philosophicus (1921), Wittgenstein 
wrote: "Philosophy is not one of the natural sciences. (The word 'philosophy' must 
mean something whose place is above or below the natural sciences, not beside them.)" 
(4.111). This conveniently left open which was superior. 

But if there is anything constant in analytic philosophy, it is change, and the oppo- 
site view of the relation between science and philosophy has dominated the second 
half of the century. Largely owing to the influence of W V Quine, many philosophers 
have come to believe that philosophy is continuous with science. Yesterday's heresy is 
today's orthodoxy. Whichever view is correct, the division between the philosophical 
analysis of concepts and the nonphilosophical scientific analysis of nonconceptual 
objects should perhaps not be taken too strictly. Concepts and hence philosophy would 
be of no use if they did not make contact with the nonconceptual world, fn addition, 
science uses concepts, many of which may be among the most fundamental of reality. 
To paraphrase Kant, perceptions without concepts are blind; concepts without percep- 
tions are empty. 

Overlapping with the latter period of logical atomism is logical positivism, which 
may be dated from Moritz Schlick's founding of the Vienna Circle in 1924. One of its 
principal doctrines was that science is a unity; and one of its principal projects was to 
show how to translate all meaningful language into scientific language, in other words, 
to reduce meaningful nonscientific language to scientific language. This project cannot 
be successful unless something distinguishes meaningful from nonmeaningful expres- 
sions. A. J. Ayer probably devoted more energy and displayed more ingenuity in trying 
to formulate a criterion of meaningfulness than anyone else. His first effort was 
presented in Language, Truth and Logic (1936), the book that became the most widely 
known statement of logical positivism and which introduced that philosophy to the 
anglophone public. The basic idea is that a sentence is meaningful if and only if it 
is either analytic (or contradictory) or empirically verifiable. Various objections were 
raised to this, and to every revision of this criterion. Part of the problem was the status 
of the criterion itself. Either it would be analytic and hence vacuous, or it would be 
empirical but then not completely confirmed. Logical positivism had been dead for some 
time when it was buried by Carl G. Hempel's "Problems and Changes in the Empiricist 
Criterion of Meaning" (1950) and W V Ouine's "Two Dogmas of Empiricism" (1951). 
Nevertheless, Ayer and others never abandoned the spirit of verifiability 

What had already begun to take the place of logical positivism in the 1940s was 
ordinary-language philosophy, one strand of which emanated from Cambridge in the 
later philosophy of Wittgenstein, the other from Oxford. One of Wittgenstein's moti- 
vating beliefs was that philosophy creates its own problems, and that means that they 
are not genuine problems at all. The confusion arises from philosophers' misuse of ordi- 
nary words. They take words out of their ordinary context, the only context in which 
they have meaning, use them philosophically, and thereby discover anomalies with the 


displaced concepts expressed by these words: "For philosophical problems arise when 
language goes on holiday." Wittgenstein questioned many of the assumptions of ana- 
lytic philosophy - from the nature and necessity of analysis to the nature of language 
- in a discursive and dialectical style so inimitable that it was as if Ludwig were talking 
to Wittgenstein. His oracular aphorisms, such as "Don't ask for the meaning, ask for 
the use" and "To understand a sentence is to understand a language" stimulated a 
variety of reactions, from the Fregean interpretations of Peter Geach and Michael 
Dummett, to the holism of Quine and Donald Davidson, to the deconstructivist 
approaches of 0. K. Bouwsma and D. Z. Phillips. 

The other strand of ordinary-language philosophy came from Oxford, under the 
leadership of Gilbert Ryle and J. L. Austin. These philosophers, more numerous than 
the Cambridge group (Antony Flew, J. 0. Urmson, and G.J. Warnock, deserve to be men- 
tioned), did not so much think that there were no philosophical problems as say that 
philosophical problems could be solved through the careful analysis of the distinctions 
inherent in ordinary language. The purpose of Austin's "Ifs and Cans" and "A Plea for 
Excuses" was to elucidate the problem of freedom and determinism, which arose from 
his understanding of Aristotle (see his Philosophical Papers, 2nd edition, p. 180). He said 
that while ordinary language was not the last word in philosophy, it was the first. He 
certainly was not opposed to philosophers developing theories. 

Austin, who had been a closet logical positivist according to A. J. Ayer, coined the 
term "performative utterance" as part of his refutation of the central thesis of logical 
positivism, namely, that all sentences that were cognitively meaningful were either true 
or false. Austin pointed out that some straightforwardly meaningful sentences, sen- 
tences that did not contain suspicious words like "beautiful," "good," or "God," were 
not the kind of sentences that could have a truth-value: "I bequeath my watch to my 
brother," "I christen this ship the Queen Elizabeth II," and "I bet ten dollars that 
Cleveland wins the pennant." Although the concept of performatives did the work it 
was designed to do, the distinction between performatives and "constatives" (roughly, 
statements) could not be sustained; and Austin replaced that distinction with another, 
between locutionary, illocutionary and perlocutionary acts. In the 1960s, John Searle, 
who was trained at Oxford by ordinary-language philosophers, showed that Austin's 
latter theory was itself inadequate and replaced it with his own fully-developed theory 
in Speech Acts (1969) and Expression and Meaning (1979). 

By the late 1960s ordinary language had lost its dominance. Some of the Oxford 
philosophers were instrumental in its demise. Searle, as mentioned, developed a full- 
fledged theory of speech acts, and then used it as inspiration for foundational work on 
the nature of intentionality and the social world. One of his teachers and a colleague 
of Austin's, H. P. Grice, developed his own theory of language use, a theory comple- 
mentary in many ways to Searle's. 

A more dramatic cause of the demise of ordinary language philosophy is attribut- 
able to one of its chief practitioners, P. F. Strawson. In Individuals (1959), he resurrected 
metaphysics, an area of philosophy that was considered unacceptable by logical posi- 
tivism. Strawson distinguished between "stipulative" (bad) metaphysics and "descrip- 
tive" (good) metaphysics. His descriptive project, to lay "bare the most general features 
of our conceptual structure," was supposed to differ from logical or conceptual analy- 
sis only "in scope and generality." At almost the same time, the American W. V. Quine 


published Word and Object (1960). His approach differed from Strawson's primarily in 
emphasizing the genesis of the most general concepts and in accommodating itself 
explicitly to empirical psychology and physics. 

Once metaphysics had been made respectable again, philosophers felt more com- 
fortable pursuing a large variety of problems in a variety of ways. Metaphysical systems 
became more elaborate when Saul Kripke used possible worlds to prove theorems about 
modal logic. Some subsequent positions can even be thought outlandish, such as David 
Lewis's view that every possible world exists, and exists in the same sense our own world 
does - outlandish but not disreputable. Some disciplines that had been relatively 
neglected between 1930 and 1960 were reinvigorated, for example, ethics and politi- 
cal philosophy by John Rawls, most notably in A Theory of Justice (1971); and some 
questions, such as the meaning of life, were mulled over by, for example, Thomas Nagel 
in an analytically respectable way. Perhaps two of the most salient characteristics of 
the period from 1970 onwards were first, the interest of analytic philosophers in the 
foundations of empirical sciences, from physics through biology to psychology, and 
second, their use of and contribution to artificial intelligence and cognitive science. 
Analysis was largely abandoned and replaced by a desire for philosophical doctrines 
that were variously more intelligible or intellectually respectable to physicists, logicians, 
or psychologists. This would explain the large presence of philosophers in cognitive 
science, linguistics, logic, and the philosophy of science; but has perhaps also led to 
what Searle has called "the rediscovery of the mind" in a book by that name. 

There were other consequences of the revival of metaphysics. Some philosophers, 
respected for their work as early as the 1950s, for example Roderick Chisholm and 
Wilfrid Sellars, but not closely associated with any of the schools we have mentioned, 
grew in significance. Some philosophers turned to the history of modern philosophy, 
notably, Strawson and Jonathan Bennett on Kant, Bennett on Locke, Berkeley and 
Hume, and Bernard Williams and Margaret Wilson on Descartes. Some philosophers 
who became important in the last quarter of the twentieth century, notably Richard 
Rorty, declared analytic philosophy misconceived, bankrupt, or similarly deficient. In 
making their position clear and in aiming at cogency, they are analytic philosophers in 
spite of themselves. 

It is likely less helpful to talk about one or another movement in philosophy after 
1965. No one method or doctrine dominated. Sometimes a philosopher championing 
a view became its most significant critic or at least moved on to something quite dif- 
ferent, paradigmatically Hilary Putnam. What can be said about the last quarter of the 
twentieth century is that the original conception of analysis and most of its presuppo- 
sitions were abandoned by almost all analytic philosophers. Gone is the assumption that 
concepts of philosophical importance are often composed of simpler sharply-defined 
concepts. Quine's arguments that there is no principled distinction between analytic 
and synthetic statements is just a special case of the broader thesis that language and 
hence thought are essentially indeterminate. 

We have been explaining and illustrating analytic philosophy in the last century 
without defining it. It probably defies definition since it is not a set of doctrines and 
not restricted in its subject matter. It is more like a method, a way of dealing with 
a problem, but in fact not one method but many that bear a family resemblance to 
each other. 


Once when Gilbert Harman was asked, "What is analytic philosophy?," he said 
(tongue firmly in cheek), "Analytic philosophy is who you have lunch with." In general, 
analytic philosophy has become highly pluralistic and in many ways hardly resembles 
what was done in the first half of the century. The refectory of analytic philosophy is 
not as clubby as it once was. Many more people sit at the table, and many more differ- 
ent kinds of food, prepared in more ways, are served. Perhaps what makes current ana- 
lytic philosophers analytic philosophers is a counterfactual: they would have done 
philosophy the way Moore, Russell, and Wittgenstein did it if they had been doing phi- 
losophy when Moore, Russell, and Wittgenstein were. The multiplicity of analytical 
styles is one reason for organizing the volume by individual philosopher and not by 

Over forty of the greatest analytic philosophers of the last century are discussed in 
this volume. At least thirty of them, we believe, would be on virtually any sensible list 
of forty outstanding analytic philosophers. Many other philosophers have almost as 
good a claim to be included in this volume. To name only some of those who are not 
alive, the following were considered and finally, reluctantly, not included: Max Black, 
Gustav Bergmann, Herbert Feigl, Paul Feyerabend, Gareth Evans, C. I. Lewis, J. L. 
Mackie, Ernest Nagel, H. H. Price, H. A. Prichard, A. N. Prior, Hans Reichenbach, Moritz 
Schlick, Gregory Vlastos, Friedrich Waismann, and John Wisdom. 

Some philosophers were excluded because they do not fit squarely within the tradi- 
tion of analytic philosophy as ordinarily understood: John Dewey, William James, 
Charles Sanders Pierce, John Cook Wilson, and, ironically, Alfred North Whitehead, 
co-author with Russell of one of the century's greatest works of logic, Principia 

While the reputations of some of the philosophers included are as high as they ever 
were, e.g. Frege and Russell, those of others have declined, not always justifiably for 
example, those of C. D. Broad and Rudolf Carnap. In making our decisions we have 
tried not to be prejudiced either for or against any school, method, or time period, but 
to reflect the relative importance of various philosophers over the entire twentieth 

We know that our selection will be controversial, even though it was influenced by 
the judgments of many colleagues. A referee of our proposal wrote that the editors 
seem to "aim at enraging the reader." Most analytic philosophers will believe that some 
other list of people would have been better. We are sympathetic. Neither of us com- 
pletely agrees with the final selection. Each believes that at least three other philoso- 
phers have a better claim to be included than some that were. In order to preserve 
"plausible deniability," we have agreed not to comment further on the lists in any 
written form, and not to appear together at any public gathering of philosophers for 
five years. 

Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

Gottlob Frege (1848-1925) 


Friedrich Ludwig Gottlob Frege was born in Wismar in 1848, and died in Bad Kleinen 
in 1925. His whole adult career was spent, from 1874 to 1918, in the Mathematics 
Department of Jena University. He devoted almost all his life to work on the borderline 
between mathematics and philosophy. In his lifetime, that work was little regarded, save 
by Bertrand Russell and Ludwig Wittgenstein; his death was marked by very few. 
Yet today he is celebrated as the founder of modern mathematical logic and as the 
grandfather of analytical philosophy. 

Frege's intellectual career was unusual. Most philosophers and mathematicians 
make contributions to diverse topics within their fields; but Frege set himself to achieve 
one particular, though extensive, task. From 1879 to 1906 he pursued this single ambi- 
tious aim: to set arithmetic upon secure foundations. Virtually all that he wrote during 
those years was devoted to this project, or to the elaboration of ideas that he developed 
in the course of trying to carry it through and took as integral to it. The term "arith- 
metic," as he used it, is to be understood in a broad sense: for him, it comprised, not 
only number theory (i.e. the theory of the natural numbers), but also analysis, that is, 
the theory of real and complex numbers. He did not attempt to construct the founda- 
tions of geometry. He viewed mathematics in the traditional way, as divided into the 
theory of quantity, and thus of cardinal numbers and of numbers measuring the mag- 
nitudes of quantities (real numbers), and the theory of space (points, lines, and planes). 
He believed these two parts of mathematics to rest on different foundations. The 
foundations of arithmetic - of number theory and analysis - are purely logical. But 
although the truths of geometry are a priori, they rest upon spatial intuition: they are 
synthetic a priori, in the Kantian trichotomy Frege accepted. Kant was therefore right 
about geometry; but he was wrong about arithmetic. All appeal to spatial or temporal 
intuition must be expelled from arithmetic: its concepts must be formulated and its basic 
principles established without recourse to intuition of any kind. 

It was to the task of establishing the purely logical foundations of arithmetic that 
Frege devoted his whole intellectual endeavor. In carrying out this task, he was led into 
some purely philosophical investigations; it is for this reason that, although a mathe- 
matician, he is now held in such high regard by philosophers of the analytic school. 
His attempt to provide arithmetic with secure logical foundations was embodied in 
three books, all of high importance, although he also published a number of articles, 


spin-offs from his central endeavor. The first of these three books was the short 
Begriffsschrift (Conceptual Notation) of 1879: this expounded a new formalization of 
logic, ft was the first work of modern mathematical logic: it contained an axiomatic 
system of predicate logic of precisely the type that was to become standard. The nota- 
tion was utterly different from that which would become standard, and was essentially 
two-dimensional, the two clauses of a conditional being written on different lines; but 
the notation was essentially isomorphic to that which Peano, Hilbert, and Russell later 
made standard. An English translation of the book and of some related articles was 
published byT W. Bynum in 1972. The Begriffsschrift received six reviews, including a 
lengthy one by Ernst Schroder; but none of the reviewers understood Frege's intention 
or his achievement. The second book was also short, though not quite so short as 
Begriffsschrift. ft was called Die Grundlagen der Arithmetik (The Foundations of Arithmetic) 
and appeared in 1884. In order that it should be accessible to as many as possible, the 
book was written without the use of logical symbols. Frege first surveys a range of rival 
theories on the status of arithmetical propositions and the nature of (cardinal) number, 
and demolishes them with trenchant arguments; this part of the book is largely philo- 
sophical in character. Then, having to his satisfaction left no space for any other theory 
but his own, he proceeds to sketch a purely logical derivation of the fundamental laws 
of arithmetic, ft is to be doubted whether any other philosophical treatise of comp- 
arable length has, since Plato, ever manifested such brilliance as Frege's Grundlagen. 
An English translation by J. L. Austin was published in 1950, and a critical edition by 
C. Thiel in 1986. This time Frege's book received five reviews, again none of them 
adequate to their subject matter. One was by Georg Cantor, who unhappily does not 
seem to have tried to understand the work, with which he might have been expected to 
be in large sympathy. 

Frege had thought that he was on the verge of success in constructing definitive 
foundations for arithmetic. He had thought that his Grundlagen would make this plain 
to the world of philosophers and mathematicians. He became intensely depressed by 
his failure to have conveyed to that world the magnitude of his achievement. At the 
same time, he became aware of deficiencies in the philosophical basis on which, in 
Grundlagen, he had rested his arguments and which underpinned his formal logic. There 
followed five years during which he published nothing, but engaged in a thoroughgo- 
ing revision of his philosophy of logic and of his formal logic. The outcome of this revi- 
sion he expounded in a lecture, Function und Begriff (Function and Concept), given in 
1891. He then set about a complete formal exposition of his foundations of arithmetic 
in his third great book, Grundgesetze der Arithmetik (Basic Laws of Arithmetic). This was 
utterly unlike Grundlagen. The first volume came out in 1893, the second volume not 
until ten years later, in 1903. The book is incomplete; a third volume must have been 
planned, but it never appeared. In the first volume of Grundgesetze, there is no argu- 
ment, only exposition. Frege began by explaining his formal logical system, and 
expounding, without giving any argument for or justification of them, the philosophi- 
cal, or, more exactly, semantic notions that underpinned it. He provides what is in effect 
a precise semantic theory for the formal language used in the book. This makes up Part 
f, of which an English translation by M. Furth was published in 1964. 

There follows in Part If a string of formal derivations, carried out in Frege's far from 
easily read symbolism, which execute in detail the program sketched in Grundlagen for 


constructing a logical foundation for the theory of cardinal numbers, including the 
natural numbers, understood as finite cardinals. Part II was concluded in the second 
volume of 1903, and is followed by a Part 111 devoted to the real numbers, atopic Frege 
had never treated in any detail before. Part III is not completed in the second volume; 
a clear intention to complete it proves that a third volume was contemplated. The first 
half of Part III is in prose, not symbols; Frege has changed his approach. He attempts 
in this half to do for real numbers what he had done in Grundlagen for cardinal 
numbers: to review and criticize rival theories so that there would appear no alterna- 
tive to his own construction of a foundation for analysis, carried out by proof in 
his formal system in the second half of Part III. Unhappily, he had lost his touch. 
Whereas in Grundlagen nothing is mentioned save what will carry the argument 
forward, the first half of Part III of Grundgesetze reads as if he was merely determined 
to get his own back on the other theorists who had neglected him, Cantor and Dedekind 
included. Criticisms are made of features of their work quite irrelevant to the main 
strand of the argument; errors are pointed out which could easily be rectified, without 
any indication of how to rectify them. A powerful critique of formalism is included. 
It is possible to extract Frege's reasons for the strategy he adopts for constructing 
foundations of analysis; but the whole lacks the brilliance and the exquisite planning 
of Grundlagen. 

When he had finished composing the second volume of Grundgesetze, Frege must 
have felt a deep contentment. Still embittered by the neglect of his work, he surely 
believed that he had attained his life's goal: he had constructed what he thought to be 
definitive foundations for the theories both of natural numbers and of real numbers. 
But, while the book was in press, he received the heaviest blow of all, delivered by one 
of his few admirers, the young Bertrand Russell. Russell wrote in June 1902 to explain 
the celebrated contradiction he had discovered in the (naive) theory of classes, and to 
point out that it could be derived in Frege's logical system (see RUSSELL). At first Frege 
was shattered. Then, as he reflected on the matter, he devised a weakening of his Basic 
Law V which governed the abstraction operator used for forming symbols for classes 
and was responsible for the contradiction; he was confident that this modification 
would restore consistency to his system. The modification was explained in an Appendix 
added to Volume II of Grundgesetze. The fact was, however, that the modified Basic Law 
V still allowed the derivation of a contradiction. Whether Frege ever discovered this is 
uncertain; but what he must have discovered was that, in its presence, none of the 
proofs he had given of crucial theorems would go through. His wife died in 1905. It 
took him until August 1906 to convince himself that his logical system could not be 
patched up; at that point he had to face the fact that the project to which he had devoted 
his life had failed. Grundgesetze remains the only part of Frege's published work of which 
no full English translation has yet appeared (save for the translation by Furth of Part I, 
which also contains the Appendix to Volume II, and excerpts in the volume of transla- 
tions by Geach and Black). 

A brief fragment among Frege's literary remains, dated August 1906, asks the ques- 
tion, "What can I regard as the outcome of my work?" In other words, "What remains 
now that the contradiction has destroyed my logical foundations for arithmetic?" 
Frege's answer was that what survived of enduring value was his logical system, 
stripped of the abstraction operator, and the whole structure of philosophical logic 


which, from 1891 onwards, had underpinned it. He continued to think about these 
topics, although it was not for years that he published anything more. Eventually, 
during the war years, he published the first three chapters of what was planned as 
a comprehensive treatise on logic, although it was never finished. In the very last years 
of his life, he finally turned again to the foundations of mathematics; reversing his 
lifelong view, he began a derivation of arithmetic from geometry, but did not carry 
it very far. 

Frege's judgment of 1906 about where the value of his work lay was at first the 
judgment of those who participated in the revived study of Frege's writings, fn the 
Preface to his Tractatus Logico-Philosophicus of 1922, Ludwig Wittgenstein had written 
of "the great works of Frege and the works of my friend Bertrand Russell," and had 
referred frequently to Frege in the book. Despite the celebrity of the Tractatus and its 
wide influence, this failed to stimulate more than a very few philosophers to find out 
anything about Frege. Rudolf Carnap, who had actually heard Frege's lectures, Alonzo 
Church, and Peter Geach continued to hold him in high regard (see CARNAP); but the 
majority of philosophers, British, German, and American, went on ignoring him. The 
revival of interest in him began in a slow way in the 1960s, and gathered momentum 
in the 1 9 70s; by the end of that decade, he had become required reading for any student 
of analytic philosophy. But the interest in his philosophy of arithmetic was meager; it 
was taken for granted that his system's having run foul of Russell's contradiction 
destroyed all its pretensions to serious consideration. Interest in Frege therefore con- 
centrated on what made him the grandfather of analytic philosophy: his philosophical 
analysis of language and of thought which underlay his formal logic. 

The key to a modern system of predicate logic is of course the quantifier-variable 
notation for generality, which Frege introduced for the first time in Begriffsschrift. He 
employed only negation, the conditional, and the universal quantifier; he did not use 
symbols for conjunction, disjunction, or the existential quantifier, but expressed these 
by means of the three logical constants for which he did have symbols. Frege insisted 
that his symbolism, unlike that in the Boolean tradition such as Schroder's, could incor- 
porate a formal language: it needed only the addition of suitable nonlogical constants, 
predicates, etc., to be capable of framing sentences on any subject matter whatever, and 
of carrying out deductive reasoning concerning it. Frege did not conceive of formulae 
in his symbolism in the way that Tarski was to do, namely as built up from atomic for- 
mulae containing free variables waiting to become bound in the process of forming 
complex formulae. Rather, he thought of them as built up out of atomic sentences. This 
required that, before attaching a quantifier, there must first be formed, from a suitable 
sentence, what we should call a predicate, but, for Frege, was a functional expression 
or expression for a concept. (He eschewed the term "predicate" as too closely associated 
with the traditional subject-predicate logic.) Such an expression was "incomplete" or 
"unsaturated": it could not stand on its own, but had gaps in it, being formed 
from a sentence by omitting one or more occurrences of a singular term. When a 
quantifier was attached to it, the bound variables governed by it were to be inserted into 
these gaps, thereby showing to what expression for a concept the quantifier had 
been attached. 

When he wanted to speak of a particular expression for a concept, Frege used the 
lower-case Greek letter i; to indicate the gaps in it; but an expression containing i; was 


no part of the formal language, but only something to be used metalinguistically 
to speak about the formal language. Thus from the sentence "Pitt respects Pitt's 
father" (which represents the way we are meant to understand the colloquial "Pitt 
respects his father") three different expressions for concepts could be formed: '% respects 
Pitt's father," "Pitt respects i;'s father," and '% respects i;'s father." This exemplifies 
the fact that a sentence can be analyzed in different ways. Frege insists in Begriffsschrift 
that these different analyses have nothing to do with the content of the sentence, but 
only with our way of looking at it; in other words, our grasp of the content of the 
sentence does not depend upon our noticing that it is possible to analyze it in one 
way or another, for example that the proper name "Pitt" occurs twice within it. In 
one sense, each of the three expressions for concepts occurs within the sentence; 
but none of the different concepts is part of the content of the sentence. By attaching 
the universal quantifier to these three expressions for concepts, we obtain respectively 
"For all a, a respects Pitt's father," "For all a, Pitt respects a's father," and "For all a, a 
respects a's father," or, colloquially, "Everyone respects Pitt's father," "Pitt respects 
everyone's father," and "Everyone respects his own father." And now, in these quanti- 
fied sentences, Frege says, the expression for the relevant concept is part of the content 
of the sentence. 

The process of forming expressions for concepts may likewise be used to form 
expressions for functions, as we ordinarily conceive them, namely by starting with a 
complex singular term within which some simpler singular term occurs. And it may 
also be used to form expressions for relations between two objects, namely by 
removing from a sentence one or more occurrences of each of two different singular 
terms. This was of importance for second-order logic, admitting quantification over 
functions and relations. 

Thus Frege's invention of the quantifier-variable notation yielded him several fun- 
damental insights. First, the conception of concept-expressions as incomplete solved 
the problem of the unity of sentences and the thoughts they express. No glue is needed 
to make the parts of the sentence adhere to one another. The concept-expression or 
relation-expression is of its nature incapable of standing alone, but can be present only 
when its argument-places are filled by singular terms to form a sentence. Or else it is 
itself the argument of a quantifier, forming a different kind of sentence: it is made to 
adhere to such terms, or to have a quantifier attached to it, and cannot exist otherwise. 
Secondly, concept-formation does not consist solely of the psychological abstraction of 
some common feature from individual objects or of the process of applying Boolean 
operations to given concepts (conjoining or disjoining them). By the process of omit- 
ting singular terms from complete sentences, or, equivalently of thinking of them as 
replaceable by other singular terms, we can arrive at expressions for concepts with new 
boundaries, and so at the concepts thus expressed. Moreover, such expressions were 
not, in general, actual parts of the sentences from which they were formed; they were, 
rather, patterns exemplified by different sentences. The expression '% respects ^'s father" 
occurs both in "Pitt respects Pitt's father" and in "Fox respects Fox's father"; the two 
sentences have, not just common words, but a common pattern. Thus we should not 
think of a sentence, or the thought it expresses, as formed out of its component parts, 
but of the components as attainable by analyzing the sentence; still less should we think 
of a concept-expression as formed out of its components, but as a result of analyzing 



a sentence. Because apprehending the possibility of analyzing a sentence in a particu- 
lar way requires us to see it as manifesting a certain pattern, which is not required for 
a simple grasp of the sentence's content, and because apprehending this possibility may 
be essential to recognizing the validity of some deductive inference, there is a creative 
ingredient in deductive inference. Such inference does not depend only upon a grasp of 
the contents of the sentences that figure in it; and this explains how deductive infer- 
ence can lead us to new knowledge, which consideration of its role in mathematics 
makes evident that it does. 

These ideas were expressed in Begriffsschrift and in Grundlagen. Parti of Begriffsschrift 
was devoted to sentential logic, and Part II to first-order predicate logic. Although Frege 
did not have the concept of the completeness of a logical system, he had in fact framed 
a complete formalization of first-order logic. Part III of Begriffsschrift is devoted to 
second-order predicate logic, involving quantification over concepts, relations, and 
functions; Frege never saw any reason for regarding the first-order fragment as espe- 
cially significant. To explain second-order quantification in the same way as first-order 
quantification, Frege has to admit the notion of an expression for a concept of second 
level, formed by removing from a sentence one or more occurrences of an expression 
for a first-level concept, or of a relational or functional expression; these second-level 
concept-expressions all have different types of incompleteness. In Part III Frege gave his 
purely logical definition of a sequence; since previously the notion of an infinite 
sequence had usually been explained in temporal terms, as involving its successive con- 
struction step by step, or else a successive diversion of attention from one term to the 
next, Frege regarded this as an essential contribution to the program of expelling intui- 
tion from arithmetic in favor of purely logical notions. It was especially important for 
number theory, since the natural numbers themselves could be defined as the terms of 
a finite sequence beginning with and proceeding from each term to its successor. 
Frege's definition of a sequence was so framed that, when the natural numbers are so 
defined, the principle of finite induction, sometimes claimed as a method of reasoning 
peculiar to arithmetic, becomes a direct consequence of the definition. Frege's 
definition of "sequence" is now generally known as the definition of the ancestral of a 
relation, namely the relation which the first term of a finite sequence has to the last 
term when each term but the last stands in the original relation to the next term. (It is 
named the "ancestral relation" because the relation "ancestor of" is the ancestral of 
the relation "parent of.") 

There are three features of Grundlagen of especial interest to philosophy in general. 
The first is the distinction between the actual (wirklich) and the objective. Frege used 
"actual" to mean "concrete" in the sense in which concrete objects are distinguished 
from abstract ones; an object is actual if it is capable of affecting the senses, directly or 
indirectly. But something may be objective even though it is not actual; an example he 
gives is the Equator. You cannot see or trip over the Equator, but it is not fictitious or 
subjective; statements about it may be objectively true or false. We can make reference 
to objects which, though objective, are not actual, and make objectively true statements 
about them. Frege thus rejected what is now called "nominalism" as based on a fun- 
damental error. This was crucial for his philosophy of arithmetic. He took numbers to 
be objects, objective but not actual: we can refer to them and make objectively true 
statements about them. 



A principle greatly stressed in Grundlagen is what has come to be known as the 
"context principle": that it is only in the context of a sentence that a word has a 
meaning. It is noteworthy that the principle is formulated linguistically, as concerning 
the meanings of words in sentences, rather than, say, as "We can think of anything 
only in the course of thinking that something holds good of it." The interpretation of 
the dictum is contentious. At the very least, it is an assertion of the primacy of sen- 
tences in the order of explanation of meaning. We must first explain what, in general, 
constitutes the meaning of a sentence, and then explain the meanings of all small 
expressions as their contributions to the meanings of sentences in which they occur. 
When we look at how Frege applies the principle in the book, it appears to have a much 
stronger significance: namely that, to secure a meaning for an expression or type of 
expression, it suffices to determine the senses of all sentences in which it occurs. Frege 
never reiterated the context principle in any subsequent writing, although there is a 
strong echo of it in Part I of Grundgesetze. 

The other salient feature is the first clear example of the linguistic turn, giving Frege 
a strong claim to be the grandfather of analytic philosophy. At a critical point of the 
book, Frege, having already argued that our notion of number is not derivable from 
sense-perception or intuition, asks, "How, then, are numbers given to us?" The ques- 
tion is both epistemological and ontological: how are we aware of numbers, and what 
guarantee is there that such objects as numbers exist? In answering it, Frege simply 
assumes that it can be equated to "How are meanings conferred on numerical terms?" 
He appeals immediately to the context principle; in virtue of this, the question reduces 
to, "What sense attaches to statements containing numerical terms?" A question about 
what objects exist and how we know of them is thus transformed into a question about 
the meanings of certain sentences. 

However, those of Frege's ideas that most interested analytic philosophers when 
interest in his work revived were the ones he expounded in his middle period 
(1891-1906). Frege had no general term for meaning, in the sense in which the 
meaning of a word or expression comprises everything that a speaker must implicitly 
know about it in order to understand it. He distinguished three features which, in this 
sense, may contribute to the meaning of a word or sentence: force, tone, and sense. The 
force of an utterance is what distinguishes an assertion from a question, and Frege rec- 
ognized only these two types of force: assertoric and interrogative. In English interroga- 
tive force is usually indicated by the inversion of the verb and subject; Frege insisted that 
the sense of a question inviting the answer "Yes" or "No" will coincide with that of the 
corresponding assertoric statement. What differentiates them is the significance of the 
utterance: in one case we ask whether the thought expressed is true, in the other we 
commit ourselves to its truth. It was important for Frege that only a complete utterance 
can carry force; a declarative sentence serving as, say, one clause of a disjunctive state- 
ment or as the antecedent of a conditional one does not have assertoric force, which is 
attached only to the statement as a whole. It was essential, Frege thought, not to con- 
strue the verb or predicate of a sentence as intrinsically containing the assertoric force 
within it. Natural languages usually lack any express means of indicating that asser- 
toric force is to be attached to a sentence, but Frege considered this a defect of them. In 
his formal language he used a symbol for just this purpose, the "judgment-stroke" (often 
called by others the "assertion sign"). It is a philosophical mistake to speak of "judg- 



merits" when all that we are concerned with is their contents; unless we are actually 
concerned with the act of recognizing them as true, we should speak in this connection 
of thoughts, rather than of judgments. Frege did not recognize imperatival or other 
kinds of force, though it may plausibly be argued that an imperative sentence expresses 
the thought that is true if the command is obeyed or the demand complied with. He 
simply declared that such a sentence expresses a command, not a thought. 

What I have called "tone," and Frege called Fdrbung (coloring) is distinguished from 
sense in that it cannot affect the truth or falsity of what is said. The English sentences 
"He has died," "He is deceased," and "He has passed away" do not differ in sense, but 
only in tone. Likewise, where A and B are sentences, the complex sentences "A and B" 
and "Not only A but B" do not differ in sense, but in tone: if either is true, the other is 
true, even if it conveys an inappropriate suggestion. The sense of a whole sentence is 
the thought that it expresses; the sense of a part of a complex expression, including a 
sentence, is part of the sense of the whole. 

In his middle period, Frege drew a distinction between the significance of an expres- 
sion and what it signifies, which he had not done in his early period. For the thing 
signified, he confusingly chose the word "Bedeutung," the ordinary German word for 
"meaning": but the Bedeutung of an expression is not part of its meaning, where 
"meaning" is understood as specified above. It is not necessary, in order to understand 
a word or phrase, to know its Bedeutung, only its sense. Frege's term is conventionally 
rendered in English either "meaning" or "reference"; neither is happy. The Bedeutung 
of a singular term is the object we use the term to talk about. It is impossible just to 
know the Bedeutung of a singular term, even if that term is logically simple, i.e. it is a 
proper name in the restricted sense (Frege misleadingly called all singular terms 
"Eigennamen" - proper names). Frege followed Kant in holding that every object of 
which we are aware is given to us in a particular way; the sense of a singular 
term embodies the particular way in which its Bedeutung is given to us in virtue of our 
understanding of the term. 

But it was not only singular terms which Frege took as having Bedeutungen: he 
ascribed them to every expression that could be a genuine constituent of a sentence, 
including incomplete ones such as concept-expressions and sentences themselves. He 
does not argue that any such expression must have a Bedeutung; he takes it for granted. 
The only question he canvasses is what kind of thing the Bedeutung of an expression of 
any given type should be taken to be. This causes much perplexity to those reading 
Frege for the first time: surely there is nothing to which a concept-expression or 
a sentence stands as a name stands to the object named. The only way to arrive at an 
understanding of Frege's notion of Bedeutung is to look at the use to which he puts 
it. That use is governed by four fundamental theses: 

1 The Bedeutung of a part of a complex expression is not part of the Bedeutung of the 

2 But the Bedeutung of the whole depends uniquely upon the Bedeutungen of its parts. 

3 If a part lacks Bedeutung, the whole lacks Bedeutung. 

4 The Bedeutung of a sentence is its truth-value - its being true or its being false. 

Thesis (1) follows from the fact that Sweden is not part of Stockholm, the capital of 
Sweden; and thesis (3) derives from the consideration that, if there is no such country 



as Ruritania, then there is no such city as the capital of Ruritania. As for the identifi- 
cation of truth-values as the Bedeutungen of sentences in accordance with thesis (4), 
that follows from Frege's extensionalist logic, despite its failure to fit natural language. 
According to it, a subsentence of a complex sentence contributes to the truth-value of 
the whole solely by its own truth-value. Counterexamples from natural language then 
have to be explained away. Instances are sentences in indirect speech following verbs 
like "said that" and "believes that"; as is well known, Frege handled these by deeming 
the sentences following "that" to have a special, indirect sense, whereby their Bedeutung 
became the senses that they would express when in direct speech. 

It is plain from these four theses that the Bedeutung of an expression constitutes its 
contribution to the determination of the truth-value of any sentence in which it occurs. 
This explains why Frege takes it for granted that any expression capable of occurring 
in a sentence without denying it a truth-value must have a Bedeutung. It also makes a 
Fregean theory of Bedeutung for a language equivalent to what we understand as a 
semantic theory for that language, which is a theory explaining how sentences of the 
language are determined as true or as false in accordance with their composition. The 
semantic value of an expression, in such a theory, is precisely that which contributes 
to the determination of the truth-value of a sentence in which that expression occurs. 
We may therefore equate the notion of the Bedeutung of an expression, as Frege con- 
ceived it, with that of its semantic value. 

In a conventional semantic theory for a formalized language, the semantic value 
of an individual constant or other singular term is an element of the domain denoted 
by the term. The semantic value of a one-place predicate is a class of elements of the 
domain; the sentence resulting from putting the term in the argument-place of the 
predicate is true if the element denoted by the term is a member of the class constitut- 
ing the semantic value of the predicate, false otherwise. Frege did not speak of the 
domain of quantification; so far as can be determined, he took the individual variables 
to range over all objects whatever, the Bedeutung of any term being such an object. Frege 
did not take the Bedeutung of a concept-expression to be a class. He called the Bedeutung 
of a concept-expression a "concept." This must not be understood in the sense in 
which we may speak of acquiring a concept or grasping a concept, which has to do 
with the senses expressed by words. In Grundlagen, the word "concept" (Begriff) had 
been used both in this way and in conformity with what was to become Frege's 
usage in his middle period; but, in that period, he took a concept to stand to a concept- 
expression as an object stands to a singular term, and thus not at all as the sense of 
that expression. 

For Frege, a concept must be distinguished from a class, which was for him a par- 
ticular kind of object. A class is the extension of a concept, comprising those objects 
that fall under the concept; but the extension of a concept is a derivative notion, only 
to be so explained. The relation of being a member of a class can be explained only as 
that of falling under a concept of which the class is the extension; any attempt to 
explain it in any other way turns the relation into that of part to whole, which is quite 
different. So we can attain the concept of a class only via that of a concept; and we can 
characterize any particular class only by citing a concept of which it is the extension. 

Since a concept-expression was for Frege incomplete, so its Bedeutung cannot be an 
object of any kind, but must be likewise incomplete, an entity needing an object to 



saturate it. This is a difficult conception, but is to be thought of by analogy with how 
we think of functions. We think of an arithmetical function as a principle according to 
which one number is arrived at, given another: not a method of arriving at the value, 
given the argument, but simply the association of the value to the argument. Its in- 
completeness consists in the fact that there is nothing to it save this association: its 
existence consists solely in its linking arguments to values. Likewise, the existence of a 
concept consists solely in its having certain objects falling under it, and others not 
falling under it. In facts, according to the doctrines of Frege's middle period, concepts 
simply are a particular type of function. For he regarded truth-values - truth and falsity 
- as themselves being objects. So a concept is a function which takes only truth-values 
as values, mapping an object falling under it on to the value true, and one not falling 
under it on to the value false. In the same way, a relation is a function with two argu- 
ments, all of whose values are truth-values. 

In his early period (1874-85), Frege had not distinguished between significance and 
thing signified; he had used the one term "content" for both without differentiation. It 
was a great advance that in his middle period he sharply distinguished them as sense 
and Bedeutung. What we tacitly know in understanding a word or expression is its sense; 
its sense is the way its Bedeutung is given to us. It is not only of an object that it holds 
good that it must be given to us in a particular way: the same holds good of concepts, 
relations, and functions. For instance, an arithmetical function may be given to us by 
means of a particular procedure for computing its value, given its argument or argu- 
ments; other procedures might serve to determine the values of just the same function. 
For this reason, the content of any piece of knowledge that we may have concerning a 
given expression can never simply consist in our knowing its Bedeutung, but must be 
our knowing its Bedeutung as given in a particular way. The Bedeutung of an expression 
is therefore no part of its meaning, where this is what we grasp in understanding the 
expression: what we grasp is its sense. We may indeed grasp more than its sense, namely 
what was called above its tone. Sense is that part of the meaning of the expression that 
is relevant to the determination of a sentence containing it as true or as false. But the 
notions of sense and Bedeutung are closely connected: again, the sense of an expression 
is the way in which its Bedeutung is given to us. (The sense is die Art des Gegebenseins, 
usually clumsily translated "the mode of presentation.") 

It is not only that each individual speaker must think of the Bedeutung of a word as 
given in some particular way, leaving it possible for different speakers to think of it as 
given in different ways. For successful communication, the speakers must know that 
the Bedeutung of a word, as each is using it, is the same. To ensure this, it must be a 
convention of the language that each associates with the word the same sense, that is, 
the same way of thinking of something as its Bedeutung. An imaginary example given 
by Frege in a letter to Jourdain is that the Bedeutung of the name 'Afla" might be given 
as the mountain visible on the northern horizon from such-and-such a place, and that 
of the name 'Ateb" as the mountain visible on the southern horizon from a certain 
other place. It may prove that the two names have as Bedeutung the very same moun- 
tain, which was not at first evident; the identity-statement 'Afla and Ateb are the same" 
is informative and reports an empirical discovery. Famously, in his celebrated essay 
"Uber Sinn und Bedeutung" of 1892, Frege used the example of the names "the Morning 
Star" and "the Evening Star," which both denote the planet Venus, to illustrate the 



distinction between sense and Bedeutung and so explain how a true statement of iden- 
tity could be informative (he had used the same example earlier in his lecture "Function 
and Concept" of 1891). 

To know the sense of an expression is, therefore, to know how its Bedeutung is deter- 
mined: not necessarily how we can determine it, since we may lack an effective means 
of doing so, but how, as it were, reality determines it in accordance with the sense we 
have given it. The sense of a part is part of the sense of the whole; the sense of any 
given expression is part of the sense of any more complex expression of which the given 
expression is part. So a grasp of the sense of an expression involves knowing how it 
may be put together with other expressions to form a complex expression - ultimately, 
a sentence - and how the sense of the complex is determined from the senses of its 
parts. To grasp the sense of a concept-expression is to apprehend a particular way of 
thinking of something incomplete as its Bedeutung, something that associates each 
arbitrary object with a truth-value: a concept that carries each object into the value 
true or the value false according as it falls under the concept or not. In general, we grasp 
the sense of a whole sentence by grasping the sense of each expression composing it, 
which is its contribution to the sense of the sentence as a whole; and to do this is to 
have a particular conception of the Bedeutung of each constituent, together with a 
grasp of how these Bedeutungen combine to yield the Bedeutung of each phrase and ulti- 
mately of the sentence itself. But the Bedeutung of a sentence is a truth-value; its sense 
Frege terms a thought. In the case of a sentence, the distinction between sense and 
Bedeutung is that between a thought and its truth-value. Thus to grasp a thought is to 
apprehend how it is determined - by reality, though not necessarily by us - as true or 
as false. And to grasp a sense that goes to compose a thought by being the sense of a 
constituent of a sentence that expresses that thought is to understand how the contri- 
bution to determining the truth-value of the thought that is made by that constituent 
is itself determined. In the words of Grundgesetze, Part I, the thought expressed by a 
sentence is the thought that the condition for its truth is fulfilled. This was an expres- 
sion of what has become the most popular form of a theory of meaning, a truth- 
conditional theory: truth is the central notion of such a theory, and meaning is to 
be explained in terms of it. 

Frege held that anyone who makes a judgment knows implicitly what truth and 
falsity are. We can express a thought without asserting or judging it to be true, which 
we do whenever we utter a sentence whose sense it is but to which assertoric force is 
not attached (e.g. when we ask whether it is true). When we judge the thought to be 
true, we "advance from the thought to the truth-value." But this advance is not a 
further thought, to the effect that the original thought is true; by prefacing the sen- 
tence expressing the thought with the words "It is true that", we do not confer asser- 
toric force on it, but merely express the very same thought as before. That is why Frege 
says, in one of his posthumously published writings, that the word "true" seems to 
make the impossible possible. Frege held the notion of truth to be indefinable: he 
rejected the correspondence theory of truth, and any other such theory that professes 
to say what truth is. 

Frege was vehemently opposed to psychologistic explanations of concepts, that is, 
of the senses of linguistic expressions. He opposed explanations in terms of the inner 
mental operations by which we acquire such concepts. The sense of any expression had 



to be explained objectively, not subjectively, in terms of the conditions for the truth of 
sentences containing the expression. A thought, for Frege, is not one of the contents 
of the mind, as is a sense-impression or a mental image. These are subjective and 
incommunicable; but it is of the essence of thoughts to be communicable. Different 
people can grasp the very same thought; it cannot therefore be a content of any of their 
minds. This rejection of psychologism was of the greatest importance: it rescued the 
philosophy of thought and of language from explanations given in terms of private 
psychological processes. Frege's alternative explanation was neither so popular nor so 
successful. He recognized no intermediate category between the subjective and the 
wholly objective. He took thoughts and their component senses to constitute a "third 
realm": like the physical universe, its inhabitants are objective, but, unlike it, they are 
not in time or space or perceived by the senses. But it is only through our grasp of the 
inhabitants of the third realm that mere sense-impressions are converted into per- 
ceptions, and so we become aware of the external world. We can grasp thoughts and 
express them: but we human beings can grasp them only as expressed in language or 
in symbolism. 

Frege's attitude to language was ambivalent. He viewed natural language as full of 
defects: only when it was conducted by means of a purified language, such as his logical 
symbolism, could deductive reasoning be confidently relied on. So some of the time he 
inveighs against language, declaring that philosophy must struggle against it and that 
his real concern is with thoughts and not with the means of their expression. Yet a 
great deal of his discussions are concerned precisely with language and its workings. 
His philosophical logic is not a theory of thought, independent of language: it is a 
systematic theory of meaning, applicable directly to a language purified of the defects 
of our everyday speech, but indirectly to natural language. The power of his theory of 
meaning rests upon the capacity of predicate logic - the logic he first invented - to 
analyze the structure of a great range of sentences and of the thoughts they express. 
Although many of his ideas were not found acceptable by later analytic philosophers, 
his theories were seen as a better model of what philosophy should aim at, in framing 
its basic theories of meaning and of thought, than anything supplied by any other 
philosopher; and his discussions of problems within that realm a better place to start 
from than any other. 

In recent years there has been a great revival of interest in Frege's philosophy of 
mathematics, the late George Boolos being one of those to have contributed greatly 
to it. The comparison between Frege's Die Grundlagen der Arithmetik and Richard 
Dedekind's Was sind und was sollen die Zahlen?, two books which approach the same 
subject matter very differently, is extremely fruitful. Dedekind is concerned to charac- 
terize the abstract structure of the sequence of natural numbers; having done so, he 
arrives at that specific sequence by an operation of psychological abstraction, a quite 
illegitimate device much favored by mathematicians and philosophers of the time. He 
acknowledges the use of the natural numbers to give the cardinality of finite classes, 
but only as a minor corollary. For Frege, by contrast, that use is central. It was for him 
the primary application of the natural numbers, and must therefore figure in their 
definition. "It is applicability alone," he wrote in Part III of Grundgesetze, "that raises 
arithmetic from the rank of a game to that of a science." He strongly opposed appeal, 
such as that made by J. S. Mill, to empirical notions having to do with one or other par- 



ticular type of application, in defining the natural numbers or the real numbers. But 
he thought it essential that, in defining them, the general principle underlying all their 
applications should be made central to their definition. Hence natural numbers were to 
be presented as finite cardinals: the operator in terms of which all numerical terms were 
to be framed was "the number of x's such that . . . x . . . ," where of course the gap was 
to be filled by an expression for a concept of first level. 

Frege's aim was to show that arithmetical were derivable from purely logical prin- 
ciples. A description of physical space as non-Euclidean is intelligible; so Euclidean 
geometry is not analytically true. By contrast, any attempt to describe a world in which 
the truths of arithmetic fail is incoherent. Since Frege characterizes logical notions as 
those which are topic-neutral, applying to things of every kind, arithmetical notions 
are already logical ones. But, like Russell's "axiom of infinity," a proposition may be 
expressed in logical terms without its truth being guaranteed by logic, ft therefore 
remains to be shown that what we take to be the fundamental truths of number theory 
are derivable from purely logical principles. 

Frege endorsed the definition of equicardinality that was becoming generally 
accepted by mathematicians, in particular Cantor: 

There are just as many Fs as Gs iff there is a relation which maps the Fs one- 
to-one on to the Gs. 

If there is a cup on every saucer on the table, and every cup on the table is on a saucer, 
we shall know that there are just as many cups as saucers on the table without neces- 
sarily knowing how many of each there are. In Grundlagen Frege enunciates a basic 
principle governing his cardinality operator: 

(*) The number of Fs = the number of Gs iff there are just as many Fs as Gs, 

"just as many as" being interpreted in accordance with the foregoing definition. He 
decides that the cardinality operator cannot be defined contextually, but requires an 
explicit definition: the one that he chooses is: 

The number of Fs = the class of concepts G such that there are just as many Fs 
as Gs. 

Here Frege appeals to the notion of a class for the first time, although he never again 
considers classes of concepts rather than of objects. But the appeal is solely for the 
purpose of framing an explicit definition of the cardinality operator; Frege uses it for 
nothing else than proving the principle (*) from it: all the theorems he goes on to prove 
about the natural numbers are derived from (*) alone, without further recourse to the 
definition of "the number of." 

The theory sketched in Grundlagen is elaborated and fully formalized in Grundgesetze, 
Part II. Grundgesetze makes extensive use of the notion of classes, or, rather, of 
Frege's generalization of it, that of value-ranges: a class is the extension of a first-level 
concept, while a value-range is the extension of a first-level function of one argument. 
The latter notion is for Frege the more fundamental one, since concepts are for 
him a special kind of function. Frege had convinced himself that the notion of a value- 
range was a logical one. The Basic Law governing the operator forming terms for value- 
ranges is Law V: 


the value-range of /= the value-range of g iff f(x) = g(x) for every x. 

It was of course this law which gave rise to the contradiction. Because of this, interest 
has centered upon a possible modification of Frege's construction of number theory, in 
which there are no value-ranges or classes, but the cardinality operator, governed 
by (*), is treated as primitive. Attention has focused on what is now called "Frege's 
Theorem," namely the proposition that, using Frege's definitions of "0," "successor," 
and "natural number," all of Peano's axioms, and hence the whole of second-order 
Peano arithmetic, can be derived in a second-order system from (*) alone. Opinions vary 
about how close this result brings us to Frege's goal of proving the truths of number 
theory to be analytic. 

While most attention has been paid to Frege's foundations for number theory, some 
has been given to his foundation for the theory of real numbers, expounded in the 
incomplete Part III of Grundgesetze. Unlike both Cantor and Dedekind, Frege does not 
first construct the rational numbers and then define real numbers in terms of them: 
true to his principle that types of number are distinguished by their applications, and 
holding that both rationals and irrationals serve to give the magnitude of a quantity, 
he simply treats rational numbers as a kind of real number, defining the latter directly. 
While cardinal numbers answer questions of the form "How many . . . ?," real numbers 
answer those of the form "How much . . . ?" Any such question that can be answered 
by a rational number can also be answered by an irrational number. There are various 
quantitative domains - lengths, durations, masses, electric charge, etc.; within each, 
the magnitude of a quantity is given as the ratio of the given quantity to some chosen 
unit quantity; these ratios are the same from domain to domain. Thus real numbers 
are to be defined as ratios of quantities belonging to the same domain; such a defini- 
tion accords with Frege's general tenet, that the definition of a type of number should 
incorporate the general principle underlying all its applications. 

In the sections of Part III included in Volume II of Grundgesetze, Frege is concerned 
to characterize quantitative domains, and he identifies them as groups of permutations 
of an underlying set satisfying certain conditions. Unknown to Frege, this work had 
been partially anticipated by Otto Holder in an article of 1 90 1 . Neither Frege nor Holder 
uses explicit group-theoretical terminology. Both of them were concerned with groups 
with an ordering upon them. Holder is generally credited with having proved the 
Archimedean law from the completeness of the ordering, which Frege also proved; but 
Frege's assumptions are much weaker than Holder's. Frege assumes only that the order- 
ing is right-invariant and that it is upper semi-linear (the ordering is linear upon the 
elements greater than any given element); Holder makes the further assumptions that 
it is also left-invariant, fully linear, and dense. This preliminary part of Frege's con- 
struction of the foundations of analysis contains substantial contributions to group 
theory, and Part III as a whole presents pregnant ideas about how real numbers should 
be explained. 

Frege's work on the philosophy of mathematics offered an explanation of how 
deductive reasoning can extend our knowledge, and a conception of the significance of 
the applications of a theory to its foundations. It also challenges us to say on what our 
recognition of mathematical truth rests, if not on pure logic or, more generally, on 
purely conceptual truths. But it offers another challenge not so often recognized. 



Frege's attempt in Part I of Grundgesetze to justify his introduction of value-ranges was 
undoubtedly a failure: he was attempting simultaneously to specify the domain of his 
individual variables and to interpret his primitive symbols over that domain. But he was 
facing a problem that is usually left untackled: how can we without circularity justify 
the existence of domains sufficiently large to contain the objects of our fundamental 
mathematical theories such as number theory and analysis? Until a convincing answer 
is given to this question, we shall not have a satisfying philosophy of mathematics. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

Bertrand Russell (1872-1970) 


Russell was the most important British philosopher of the twentieth century. At the 
start of the century he helped to develop the new theories that transformed the study 
of logic at this time, but his greatest contribution was not to logic itself. Instead it lay 
in developing and demonstrating the philosophical importance of this new logic and 
thereby creating his "logical-analytic method," which is the basis of the analytical style 
of philosophy as we know it today. The result is that we can still read Russell's writings 
as contributions to contemporary debates. He is not yet someone whose works belong 
only to the history of philosophy. He lived to be nearly 100 and there is every reason 
to expect that some of his writings will have an active life and age at least as great as 
his. A classic instance is provided by his introduction to philosophy, The Problems of 
Philosophy (his "shilling shocker" as he liked to call it), which, though published in 
1912, remains one of the best popular introductions to the subject. 

Early life 

Despite the fact that as a philosopher Russell remains almost a contemporary, in other 
respects his life now seems very distant from us. His family, the Russells, was one of the 
great Liberal families of British politics: his paternal grandfather, Lord John Russell, had 
been Prime Minister twice during the first half of the nineteenth century, and Russell 
describes meeting Mr. Gladstone several times. His parents, Viscount Amberley and his 
wife Kate, were friends with John Stuart Mill, who agreed to act as an honorary god- 
father to their young son Bertrand. Mill in fact died during the following year, too soon 
for Bertrand to make his acquaintance. Much more traumatic for the young child, 
however, was the death of both his parents soon afterwards, so that in 1 8 76 he was left 
at the age of 4 in the care of his grandparents, indeed of just his grandmother after 
his grandfather's death in 1878. Russell described his lonely childhood in his 
Autobiography. His rebellious elder brother Frank was sent away to school, but Bertrand 
("a solemn little boy in a blue velvet suit," 1967: 30) was educated at home, brought 
up in a constricting atmosphere whose narrow limits were fixed by his grandmother's 
strict Presbyterian beliefs. The only refuge that the young Bertrand found was in the 
privacy of his own thoughts; he kept a secret diary in code in which he set down 
his growing doubts about religious orthodoxy. There can be little doubt that Russell's 



troubled later emotional life (he had four marriages) was affected by this childhood. In 
his writings there are many passages in which he refers indirectly to it; for example, 
when writing in 1916 about the difficulties of marriage, he remarks that "The funda- 
mental loneliness into which we are born remains untouched, and the hunger for inner 
companionship remains unappeased" (1916: 191). 

From an early age his precocious talent in mathematics had been recognized, and 
in 1890 he went to Trinity College, Cambridge to study mathematics. Despite his delight 
in escaping from his grandmother, however, he soon found himself dissatisfied with the 
antiquated teaching of mathematics at Cambridge. So in 1893 he switched to the study 
of philosophy, a subject into which he had been initiated through membership of the 
Cambridge 'Apostles" (a private society largely dedicated to the discussion of philoso- 
phy) and friendship with the philosopher J. M. E. McTaggart, then a young Fellow of 
Trinity. In 1894 he obtained a first class result in his final examinations and almost 
immediately started work on a dissertation in the hope of winning a prize fellowship at 
Trinity College. At the same time he married, and then travelled with his first wife, Alys, 
to Germany. In his Autobiography he recounts a moment of clear-minded future reso- 
lution during this honeymoon: 

During this time my intellectual ambitions were taking shape. I resolved not to adopt a pro- 
fession, but to devote myself to writing. I remember a cold, bright day in early spring when 
I walked by myself in the Tiergarten, and made projects of future work. I thought that I 
would write one series of books on the philosophy of the sciences from pure mathematics 
to physiology, and another series of books on social questions. I hoped that the two series 
might ultimately meet in a synthesis at once scientific and practical. My scheme was 
largely inspired by Hegelian ideas. Nevertheless, I have to some extent followed it in later 
years, as much at any rate as could have been expected. The moment was an important 
and formative one as regards my purposes. (1967: 125) 

This passage (though written with the benefit of hindsight) is remarkably prophetic; 
Russell had no settled profession (he held teaching positions for only about ten years) 
and for most of his life he made his living by writing, in which he had remarkable pro- 
ficiency. He wrote about seventy books, which do indeed form two series: there are about 
twenty books of philosophy, mostly on "the philosophy of the sciences"; and many of 
the rest concern "social questions," though it cannot be said that the two series meet 
in a synthesis. His most popular book was A History of Western Philosophy, despite the 
fact that this is an unreliable and distinctly idiosyncratic book in which Russell devotes 
most space to ancient and medieval philosophy. 

Breaking with idealism 

During the 1890s the dominant school of philosophy in Cambridge, as in Britain gen- 
erally, was idealist. Under McTaggart's influence Russell chose to work within a broadly 
idealist framework (as the Tiergarten testament quoted above shows), and the project 
he selected for his fellowship dissertation was that of providing a revised a priori foun- 
dation for geometry, one that would take account of the possibility of non-Euclidean 
geometries in a way that Kant's famous account does not. Russell argued that Kant's 



conception of the conditions of the possibility of experience had been too limited. What 
is important, and thus a priori, is that space be of a constant curvature, but it is not an 
a priori matter just what its curvature is - for example whether it is zero, as Euclid main- 
tained, or positive, as Riemann proposed. Russell was duly elected to a six-year prize fel- 
lowship at Trinity College in 1895 and in 1897 he published a revised version of the 
dissertation, An Essay on the Foundations of Geometry. 

As well as offering a Kantian foundation for geometry Russell argued in a Hegelian 
fashion that within the abstract conception of points in space characteristic of 
geometry there are "contradictions," which can only be resolved by incorporating 
geometry into an account of the physical structure of space. This led him into a study 
of the foundations of physics and in particular to a study of the problems associated 
with the continuity of space and time. He initially approached these matters with the 
presumption that these problems arise from a conflict between, on the one hand, the 
fact that individual points or instants differ only in respect of their relations and, on 
the other, the requirement that "all relations are internal," which he took to imply that 
differences in the relationships between things are dependent upon other differences 
between these things. Russell summed up the conflict here as "the contradiction of 
relativity": "the contradiction of a difference between two terms, without a difference 
in the conceptions applicable to them" ("Analysis of Mathematical Reasoning," Papers, 
2: 166). The presumption that there is a contradiction here was a commonplace among 
the idealist logicians of the period, such as F. H. Bradley, and was central to their denial 
that there are any relational truths and thus to their metaphysical monism, ft was 
therefore by calling this presumption into question that Russell made his break with 
idealism. The key to this was his affirmation of the independent reality of relations, 
which he proposed in his 1899 paper "The Classification of Relations" (see Papers, 2). 
Once this move was made, the alleged "contradiction of relativity" is dissipated and 
Russell was free to approach the issues raised by the continuity of space and time afresh. 

Although Russell's papers from this period show him finding his own way to this 
anti-idealist thesis, he always acknowledged the decisive importance of G. E. Moore's 
writings at this time ("It was towards the end of 1898 that Moore and I rebelled 
against both Kant and Hegel. Moore led the way, but I followed closely in his footsteps," 
1995a: 42). G. E. Moore was two years younger than Russell. Having been drawn 
from the study of classics to that of philosophy at Trinity College partly through 
Russell's influence he graduated in 1896 and completed his own, successful, disserta- 
tion for a prize fellowship in 1898. It is in this dissertation that Moore works out 
his own break with idealism. His basic claim is that of the unqualified reality of the 
objects of thought, propositions, as things is in no way dependent upon being thought 
about. Moore further maintained that there is no reason to duplicate ontological struc- 
tures by hypothesizing the existence of facts for true propositions to correspond to. 
Instead there are just propositions and their constituents: the world just comprises 
the totality of true propositions, and an account of the structure of propositions is an 
account of the structure of reality itself. One implication of this is that the structure 
of space is independent of our experience of it and, therefore, of the conditions 
under which experience of it is possible. So Moore was very critical of Russell's neo- 
Kantian account of geometry, and Russell quickly came to agree with Moore on this 
matter (see MOORE). 



Russell read Moore's dissertation at an early stage, and immediately accepted many 
of Moore's central points, including in particular his conception of a proposition. As 
we shall see, many of his later difficulties can be traced back to this. But at the time 
Russell was exhilarated by the possibilities that this new realist philosophy opened out 
before him: 

But it was not only these rather dry, logical doctrines [concerning the reality of relations] 
that made me rejoice in the new philosophy. I felt it, in fact, as a great liberation, as if I 
had escaped from a hot-house on to a wind-swept headland. (1995a: 48) 

The principles of mathematics 

This tremendous sense of intellectual liberation quickly became focused on a new 
project, which was to dominate Russell's thought and life for the next ten years: the 
"logicist" project of demonstrating that "all mathematics is Symbolic Logic." The occa- 
sion which fired Russell's enthusiasm for undertaking this project was his visit in July 
1900 to the International Congress of Philosophy in Paris, where he heard Peano 
discuss his formalization of arithmetic using new logical techniques. Peano did not 
himself seek to provide a purely logical foundation for mathematics; he did not offer 
logical definitions of the concepts "0," "successor," and "number" which occur in his 
postulates. But, on hearing him, Russell jumped to the hypothesis that definitions of 
this kind should be possible, and thus that mathematics is, in the end, just logic. In 
making this jump Russell was drawing on his recent close study of the philosophy of 
Leibniz, which, fortuitously, he had undertaken the year before (simply because he 
stood in for McTaggart who should have been teaching it). Russell recognized that 
Leibniz had also conceived this hypothesis but had been prevented from demonstrating 
it, largely because of the inadequacies of the traditional logic to which he adhered. But 
with the richer resources of the logic employed by Peano (which Russell immediately 
used to develop a new logic of relations), Russell supposed that Leibniz's logicist hypoth- 
esis could now be vindicated. 

An important aspect of Russell's new project was the opportunity it provided him to 
continue his criticisms of idealist philosophy: 

The questions of chief importance to us, as regards the Kantian theory, are two, namely, 

(1) are the reasonings in mathematics in any way different from those of Formal Logic? 

(2) are there any contradictions in the notions of time and space? If these two pillars of 
the Kantian edifice can be pulled down, we shall have successfully played the part of 
Samson towards his disciples. (1903: 457) 

In providing a negative answer to the second of these questions Russell drew on the 
work of the great German mathematicians of the nineteenth century, Dedekind, 
Weierstrass, and Cantor, whose work he had discovered a few years earlier but had not 
then appreciated fully because of his attachment to idealist doctrines. He now devoted 
a central section of his new book, The Principles of Mathematics, to a careful exposition 
of their philosophy of the infinite, from which he concluded "that all the usual 
arguments, both as to infinity and as to continuity, are fallacious, and that no definite 



contradiction can be proved concerning either" (1903: 368). Since the idealist logicians 
had advanced the opposite view, and then used the infinity of space and time to argue 
for their unreality, Russell felt that he was here providing a definitive refutation of their 

In laying out this new philosophy Russell worked at extraordinary speed. He started 
writing The Principles of Mathematics in October 1900 and by the end of the year he 
had completed a first draft: the book as we now have it is more than 500 pages long 
and 300 of these come unchanged from that first draft. This period was, he wrote, 

an intellectual honeymoon such as I have never experienced before or since. Every day 
I found myself understanding something that I had not understood on the previous day. 
I thought all difficulties were solved, all problems were at an end. (1995a: 56) 

But, he continues, 

The honeymoon could not last, and early in the following year intellectual sorrow 
descended upon me in full measure. (1995a: 56) 

The main reason for the onset of this sorrow was his discovery, early in 1901, of 
"the contradiction," now usually known as "Russell's paradox." This is a contradiction 
that can be easily demonstrated just at the point at which one seeks to develop 
elementary logic into set theory in order to show how arithmetic can be established 
on the basis of logic alone. Russell discovered the contradiction when reflecting upon 
Cantor's "paradox" that there is no greatest cardinal number. Cantor's paradox rests 
on the theorem that the number of subsets of a given set S is always greater than the 
number of members of S itself. Cantor proves this theorem by deducing a contradiction 
from the hypothesis that there is a one-to-one correlation between the subsets of S and 
the members of S, which would imply, on the contrary, that their numbers are the same. 
The contradiction arises as follows: consider that subset of S whose members are just 
those members of S which do not belong to the subset of S with which they are corre- 
lated under the hypothesized correlation. Since this so-called "diagonal" set D is a subset 
of S it too must be correlated with a member of S, say d. Cantor now asks whether d 
belongs to D: given the way d and D have been defined, it turns out that d belongs to D 
if and only if d does not belong to D, from which it is easy to derive the explicit contra- 
diction that d both belongs to D and does not belong to D. 

The step from Cantor's theorem to Russell's paradox is very simple. Instead of 
Cantor's hypothetical one-to-one correlation between the members of a set and 
its subsets, consider instead the non-hypothetical identity relation between anything 
and itself. Then the analogue of Cantor's "diagonal" set D of things that are not 
members of the set with which they are hypothetically correlated becomes simply 
the set R of things that are not members of themselves. Under the "identity" correla- 
tion R is of course "correlated" with itself; hence in asking whether R belongs to that 
with which it is correlated we are simply asking whether R belongs to R. But since R 
just is the set of things which do not belong to themselves, it follows that R belongs 
to R if and only if R does not belong to R. This too immediately gives rise to an explicit 
contradiction, but in this case the derivation does not depend on a hypothetical 



correlation which is thereby proved not to exist, but only on the non-hypothetical 
identity of a thing with itself which cannot be rejected. So in this case there is no 
obvious positive conclusion to be drawn comparable to Cantor's theorem. There is 
instead the utterly dismaying implication that there is something seriously amiss in the 
foundations of logic. 

As soon as he had discovered this contradiction Russell communicated it to Frege, 
whose works he had just recently read properly for the first time and recognized for 
what they were, namely much the most sophisticated attempt to develop a logicist 
program of the kind he was also engaged upon. Frege received Russell's letter just as 
the second volume of his Grundgesetze was in press, and added the famous Appendix II, 
which begins, "Hardly anything more unfortunate can befall a scientific writer than to 
have one of the foundations of his edifice shaken after the work is finished." Frege did 
in fact suggest a way of circumventing Russell's paradox, but he could not show that 
it worked, and it is now known not to. Russell also attempted to find a way around 
the paradox and ended The Principles of Mathematics with a tentative proposal that the 
definition of the set R is ill-formed because a set, being of a different "type" from 
that of its members, cannot be a member of itself. As we shall see below, this is a pro- 
posal to which he returned later; but in the context of the The Principles of Mathematics 
it was not easy for him to advance it, since it conflicts with the conception of logic 
advanced there, namely that the truths of logic are truths which are absolutely 
universal, and are therefore not to be restricted by considerations arising from the type 
of thing under discussion. 

The contradiction was not the only problem to delay publication of The Principles of 
Mathematics until 1903 and to dominate his research for the next few years. He also 
ran into a tangle of difficulties concerning the structure of judgment, with which he 
continued to grapple thereafter. These difficulties are, broadly, of two kinds: concern- 
ing (1) the unity of judgment and (2) the structure of general judgments. 

In The Principles of Mathematics Russell's treatment of these matters is expressed 
through a discussion of the structure of propositions, which, following Moore, he takes 
to comprise both the objects of judgment and the objective structure of the world. So 
conceived, propositions are not representations which, when true, correspond to a fact. 
Instead true propositions just are facts - there is no difference between the death of 
Caesar and the (true) proposition that Caesar is dead. Since his death is something that 
befell Caesar himself, Caesar is himself a "constituent" of the proposition that Caesar 
is dead. Indeed the proposition just is a "complex" whose constituents are Caesar and 
death (which is a "predicate"). 

The difficulty that now arises concerning the "unity" of judgment is that of 
explaining how it is that a complete proposition is constituted. Russell discusses this in 
connection with the proposition that A differs from B. The constituents here are A, B, 
and difference; but specifying them does not yet specify the proposition in question, for 
they are also the constituents of the different proposition that B differs from A 
(because difference is a symmetric relation this is actually a poor case to have taken. 
The point is much clearer with an asymmetric relation, such as occurs in the 
proposition that A is larger than B, which is manifestly different from the propo- 
sition that B is larger than A although it has the same constituents). Russell sums up 
his discussion: 



The difference which occurs in the proposition actually relates A and B, whereas the 
difference after analysis is a notion which has no connection with A and B. (1903: 49) 

Furthermore, Russell notes, it does not help if one adds that the difference in the case 
we want is a difference of A from B, for all that these additions do is to add further rela- 
tions to the supposed constituents of the proposition without "actually relating" the 
relation of difference to the right terms. The problem is that 

A proposition, in fact, is essentially a unity, and when analysis has destroyed the unity, no 
enumeration of constituents will restore the proposition. (1903: 50) 

The point raised here is one that was to come back to plague him. In The Principles 
of Mathematics Russell, having identified it, simply sets it aside for further treatment. 
This is disappointing, for here there is a straightforward challenge to the Moore-Russell 
conception of a proposition as a "complex whole" comprised of its elementary con- 
stituents. Indeed the point was not new: this difficulty concerning relational judgments 
had been famously set out by F. H. Bradley in chapter III of Appearance and Reality 
(1893). The person who first saw clearly the way to defuse the issue here wasFrege: for 
through his famous "context" principle (presented in his Grundlagen, 1884) that it is 
only in the context of a sentence that a word has a meaning, he acknowledges the irre- 
ducible primacy of judgments instead of regarding them as "complexes" to be con- 
structed out of elementary constituents. We shall see below that Russell's theory of 
descriptions includes a partial acknowledgment of the context principle; but he himself 
never generalizes it to solve this problem of the unity of judgment. 

The other general type of difficulty that Russell encountered in The Principles of 
Mathematics concerns the structure of general propositions such as the proposition that 
I met a man. The difficulty here is supposed to come from the fact that, on the one hand, 
the concept a man is a constituent of this proposition; but, on the other hand, such a 

does not walk the streets, but lives in the shadowy limbo of the logic-books. What I met 
was a thing, not a concept, an actual man with a tailor and a bank-account or a public 
house and a drunken wife. (1903: 53) 

Russell's argument here is intuitive and questionable. But it is clear that the tension 
arises from the dual role of propositions as both objects of thought (and thus consti- 
tuted of concepts such as a man) and situations within the world (and therefore con- 
stituted not from general concepts, but from particular men). 

Russell's way of resolving this tension is to say that concepts such as a man "denote" 
the things that a proposition in which they occur is "about"; and it is the things that 
are denoted in this way that "walk the streets" etc. It is difficult at first not to interpret 
this talk of that which a proposition is "about" as a way of implicitly specifying a "truth- 
maker" for the proposition distinct from the proposition itself; but this of course would 
be entirely inimical to Russell's conception of a proposition. Furthermore, Russell 
develops his account of denoting in such a way as to make this interpretation 
inappropriate. For he goes on to argue that 



some man must not be regarded as actually denoting Smith and actually denoting Brown 
and so on: the whole procession of human beings throughout the ages is always relevant 
to every proposition in which some man occurs, and what is denoted is essentially not each 
separate man, but a kind of combination of all men. (1903: 62) 

As Russell acknowledges, such a "combination of all men" is a very paradoxical object; 
in the case of a man, what is denoted is supposed to be Smith or Brown or . . . (for the 
whole human race). 

It is in fact clear enough what Russell is seeking to do here: namely, to effect a reduc- 
tion of general propositions to propositions that involve only disjunctions or conjunc- 
tions of singular propositions, for which the tension between propositions as objects of 
thought and as truthmakers is not so acute. He writes that "the notion of denoting may 
be obtained by a kind of logical genesis from subject-predicate propositions" (1903: 54) , 
and as Geach has observed, Russell's account is in this respect comparable to medieval 
theories of suppositio. But Russell muddles things by supposing that he needs to hold 
that there are disjunctive and conjunctive combinations of things denoted by the denot- 
ing concepts that occur in the general propositions. These are the "paradoxical objects" 
which the propositions in question are to be "about." Not only are such objects intrin- 
sically objectionable (as Russell himself acknowledges, 1903: 55n.), this way of coming 
at the matter makes it impossible to provide a coherent treatment of propositions 
involving multiple generality, since there is no way of representing scope distinctions, 
such as the distinction between the two ways of interpreting "Everyone loves someone." 

Russell in effect acknowledges this point himself when discussing variables, both free 
and bound. He would like to handle free variables within his theory of denoting as cases 
of the denoting concept any term; but he can see that this approach does not deal prop- 
erly with the role of repeated variables. His discussion of bound variables occurs as part 
of his exposition of Peano's conception of "formal implication" as a universally quan- 
tified conditional, as in "for all x, if x is a man then x is mortal." Russell wants to be 
able to apply his theory of denoting concepts to the quantifier "for all x," as a concept 
denoting some combination of things which the whole proposition is "about." But he 
can also see that where there are multiple quantifiers binding different variables (as in 
both interpretations of "Everyone loves someone") the variables are tied to the quan- 
tifiers in a way that blocks off this conception of the denotation of a quantifier. So 
although he cannot bring himself to say so explicitly, his theory of denoting concepts 
is inadequate to the new logic of quantifiers and variables upon which the logicist 
project of The Principles of Mathematics is founded. 

The theory of descriptions 

Two years later, in 1905, Russell published his most famous paper, "On Denoting." He 
begins by, in effect, developing his earlier discussion of formal implication into a sys- 
tematic account of the propositions expressed by sentences involving what he now calls 
"denoting phrases" such as "all men," "a man," and "no man." There is now no talk 
of denoting concepts; instead he uses the universal quantifier and bound variables to 
specify the propositions expressed by these sentences by reference to the truth of simpler 
propositions. Thus he now says that the proposition expressed by "I met a man" is the 



proposition that propositions of the type "I met x, and x is human" are not always false 
(Papers, 4: 416). 

Clearly, much is here assumed, for example the interdefinability of the existential and 
universal quantifiers. There is also a degree of oversimplification, since, as he recog- 
nizes when dealing with multiple quantifiers, he actually needs to specify the variable 
in the quantifier; in the case above he should have said "are not always false of x." 
Setting these points aside, what is worth considering is whether Russell offers any 
general account of quantifiers and variables to replace the theory of denoting concepts 
that has been tacitly discarded. In "On Denoting" itself Russell is unhelpful: he just says 
"Here the notion 'C(x) is always true' is taken as ultimate and indefinable" (Papers, 4: 
416). If, however, one looks ahead to the discussion of the universal quantifier in the 
introduction that Russell wrote to Principia Mathematica (1910), one finds him using 
the language of "ambiguous denotation" to sketch what is now recognizable as a sub- 
stitutional account of the quantifier. For he now says that we assert a universal propo- 
sition in order to condense the assertion of all the substitution instances "ambiguously 
denoted" (1910: 40) by the propositional function that occurs in our universal propo- 
sition; and the truth of the universal proposition depends on the "elementary truth" of 
all these substitution instances (p. 42). Since the motivation behind the original theory 
of denoting concepts was that propositions involving all men, a man, etc. are in some 
way "about" their instances, it is unsurprising that he ends up with a substitutional 
treatment of quantification, dealing, of course, with substitutions in propositions, not 

Russell's main topic in "On Denoting" is the structure of propositions whose expres- 
sion involves definite descriptions, phrases such as "The present President of the USA." 
In The Principles of Mathematics he had applied his theory of denoting concepts to such 
propositions in order to provide an account of why it is that true identities, propositions 
expressed by sentences such as "Bill Clinton is the present President of the USA," are 
of interest to us (1903: 62-4). The problem here is familiar: if we just take it that we 
have two names for the same thing, and that the proposition's constituents are just the 
thing thus named twice and the relation of identity, it seems that the very same propo- 
sition is also expressed by "Bill Clinton is Bill Clinton," which is of no interest to us. In 
The Principles of Mathematics Russell took it that this problem is solved by the hypoth- 
esis that the description "the present President of the USA' introduces a corresponding 
denoting concept into the proposition expressed through its use, which of course does 
not occur as a constituent of the proposition expressed by "Bill Clinton is Bill Clinton." 

Russell does not explain how this denoting concept occurs as a constituent of the 
proposition expressed, and he says other things about it on the basis of which it is easy 
to reopen the old problem. Since the proposition cannot comprise Bill Clinton's identity 
with the denoting concept in question, it seems that it must comprise Bill Clinton's iden- 
tity with the thing denoted by the denoting concept, that which the proposition is 
"about" in Russell's intuitive sense. But this thing is of course just Bill Clinton himself, 
and we have now come back to the difficulty of showing why this proposition is of any 
interest to us since it is equally expressed by "Bill Clinton is Bill Clinton." If one looks 
to Russell's extensive unpublished writings on this matter from the period 1903-5 (see 
Papers, 4), one can, I think, see him identifying this difficulty for his old position. He 
also begins to think about a different issue, also problematic for his old position, which 



arises from his critical reaction to some works by the contemporary Austrian philoso- 
pher Alexius Meinong, which he studied closely at this time. 

In this case the issue concerns the proper treatment of "empty" descriptions, descrip- 
tions such as "the present King of France" which, though meaningful, describe nothing 
that actually exists. Russell interpreted Meinong as advancing a theory of "objects" 
according to which empty descriptions of all kinds describe an object, even descriptions 
of impossible objects such as "the round square." In "On Denoting" and thereafter 
Russell ridiculed this position as conflicting with "the robust sense of reality" which 
"ought to be preserved even in the most abstract studies" (1919: 169-70); butin truth 
Meinong's position was a good deal more subtle than Russell appreciated. What is 
important here, however, is that through thinking about Meinong's work Russell came 
to appreciate the importance of constructing a theory that would allow for the possi- 
bility of meaningful sentences that include empty descriptions - sentences such as "The 
present King of France is bald." Since Russell took it that the meaning of a name was 
just the name's bearer, it followed at once that such descriptions were not names. This, 
he recognized, did not rule out the treatment of such descriptions as introducing denot- 
ing concepts into a proposition. But, he argued, there was still a problem: the proposi- 
tion expressed by "The present King of France is bald" ought to be about the present 
King of France. But there is no such thing; so the theory implies that the proposition is 
about nothing, which Russell takes to imply that "it ought to be nonsense." But, he con- 
tinues, "it is not nonsense, since it is plainly false" ("On Denoting," Papers, 4: 419). 

This argument is condensed, but the way to understand the crucial move from being 
"about nothing" to "being nonsense" is to connect the "aboutness" thesis with the thesis 
that the truth or falsehood of a proposition with a denoting concept depends on the 
truth or falsehood of the propositions specified by reference to that which the first 
proposition is about. For in the light of this, a proposition about nothing will be one for 
which there are no such propositions to determine its truth or falsehood. So it can be 
neither true nor false itself; but this is absurd since propositions are inherently either 
true or false. Hence it follows that the sentence with an empty description fails after all 
to express a proposition: in which case "it ought to be nonsense." 

We now have the two "puzzles" which, Russell says in "On Denoting," an adequate 
theory of descriptions must solve: (1) Why is it that questions about identity are often 
of interest to us - how can it be that George IV wished to know whether Scott was the 
author of Waverley but did not wish to know that Scott was Scott? (2) What proposi- 
tions are expressed by sentences with empty descriptions, such as "The present King of 
France is bald," and how is their truth or falsehood determined? Before showing how 
his new theory can solve these puzzles, Russell explains why he rejects the view that 
one should account for these puzzles by distinguishing between the "meaning" and the 
"denotation" of a description. He specifically mentions Frege in connection with this 
view and it is natural to think of him as arguing here against Frege's theory of Sinn 
and Bedeutung which was of course precisely introduced to handle the first of Russell's 
puzzles and seems well suited to handle the second (see FREGE). 

Russell's argument in "On Denoting" against the "Fregean" position is notoriously 
obscure. His conclusion is that there is an "inextricable tangle" in the account this posi- 
tion offers of the relation between the meaning and the denotation of a description 



(Papers, 4: 422); but in truth it is Russell's own discussion which certainly appears to 
be an inextricable tangle. My own view is that Russell's argument is vitiated by the fact 
that he adapts Frege's theory to his own different conception of a proposition, and 
thereby crucially distorts Frege's actual position. (Russell himself acknowledges the dif- 
ferences here: Papers, 4: 419 n. 9.) This can be illustrated by considering a slightly later, 
and considerably clearer, discussion by Russell of the same issue, in "Knowledge by 
Acquaintance and Knowledge by Description, " in which he is considering the structure 
of the proposition expressed by "the author of Waverley is the author of Marmion" 
(Papers, 6: 159). Russell argues here that (1) the proposition involves an identity; but 
(2) plainly does not involve the identity of the meanings of "the author of Waverley" 
and "the author of Marmion"; so (3) it must comprise the identity of the denotations 
determined by the meanings of "the author of Waverley" and "the author of Marmion." 
But these denotations are just Scott himself, so the proposition in question is just the 
proposition that Scott is Scott. Yet the whole point of the meaning/denotation theory 
was to preserve the distinction between interesting and trivial identities. 

For Frege this argument simply gets off on the wrong foot. The thought expressed 
by "the author of Waverley is the author of Marmion" is indeed a thought about iden- 
tity; but the thought itself is not structured by the relation of identity, but by the sense 
employed here of this relation, and this sense can then relate the senses expressed by 
the two descriptions without turning it into a thought that these senses are the same. 
Russell's argument against the meaning/denotation distinction as applied to descrip- 
tions only works insofar as it depends on a non-Fregean hybrid conception of a 
proposition which violates Frege's distinction between Sinn and Bedeutung. 

The fact that Russell's argument against Frege is a failure does not, of course, vin- 
dicate Frege's treatment of descriptions as functional expressions or undermine 
Russell's own theory of descriptions. The key to this new theory is, as Russell put it later, 
that descriptions are "incomplete symbols," that is, phrases which "have no meaning 
in isolation" but are only "defined in certain contexts" (1910: 66). As such, for Russell, 
descriptions such as "the man" are to be regarded as essentially similar to the other 
denoting phrases discussed in "On Denoting," e.g. "a man" and "all men." They have 
"no meaning in isolation" in the sense that there is no thing (not even a concept) that 
is their "meaning" and that occurs as a constituent of the propositions expressed by 
sentences in which they occur. Instead they contribute to these propositions in more 
complex ways by fixing their structure, in the way that Russell conceives of the role of 
the universal quantifier as described above. It is therefore no surprise that on Russell's 
new theory of descriptions, the role of descriptions is elucidated by spelling out the 
quantificational structure of the propositions expressed by sentences in which they 
occur. This turns out to be a complex matter since Russell sticks to his initial assump- 
tion that all such propositions involve only the universal quantifier. But we can grasp 
the key points of Russell's new theory by providing just the first stage of it, which is 
that the proposition expressed by "The author of Waverley is Scott" is more clearly iden- 
tified as the proposition expressed by the following sentence: 

For some x, (i) x is an author of Waverley and, (ii) for all y, if y is an author of Waverley 
then y = x, and, (iii) x = Scott. 



Once the proposition is identified in this way solutions to Russell's two puzzles are 
immediate. There is no reason to identify the complex proposition thus expressed with 
the proposition that Scott is Scott, for the description "the author of Waverley" is not 
construed as giving rise to a complex name of a constituent within the proposition 
which turns out to be just Scott. So the distinction between significant and trivial iden- 
tities is clearly preserved. Secondly, the role of empty descriptions is easily allowed for: 
the proposition expressed by "The present King of France is bald" is to be identified as 
that expressed by the sentence: 

For some x, (i) x is a present King of France and, (ii) for all y, if y is a present King 
of France then y = x, and, (iii) x is bald. 

It is then unproblematic that there is such a proposition, and that it is "clearly false" as 
Russell declares that it should be. 

The fact that these solutions are so straightforward, and do not depend on the under- 
lying theory of propositions (though they are consistent with it), shows that one can 
abstract Russell's theory of descriptions from this underlying theory. This is in fact what 
has largely happened to Russell's theory of descriptions: it is taken to rest on the thesis 
that descriptions are quantifiers, and as such the Russellian position is best conceived 
of as one that employs a restricted "definite" quantifier to construe definite descriptions, 
so that the logical form of "the author of Waverley is Scott" can be better captured by 
construing it as 

For the x who is an author of Waverley, x = Scott. 

One can then interpret Russell's reduction of the definite quantifier to other quantifiers 
as a misleadingly expressed way of spelling out the truth-conditions of this sentence. 
Once the matter is handled in this way, the debate with Frege can be re-opened, as a 
debate as to whether this way of construing descriptions is preferable to Frege's 
approach, according to which they are complex singular terms, comparable to func- 
tional expressions in mathematics, such that the logical form of "the author of Waverley 
was Scott" is captured by construing it as 

The author of (Waverley) = Scott. 

This debate was famously revived by Strawson in his attack on Russell. Strawson 
introduced a range of linguistic data concerning the use of empty descriptions in 
situations which conflict with our "presupposition" that descriptions are non-empty, 
and argued on their basis that a Fregean position is in fact preferable to Russell's; in 
many cases, Strawson argued, we take it that the use of empty descriptions issues in 
statements that are "neither true nor false" and not "plainly false" in the way that 
Russell maintained. In fact, however, the linguistic data in this area are indeterminate, 
and most contemporary discussions focus instead on the relative merits of alternative 
accounts of the logical behavior of descriptions in complex sentences involving 
temporal modifiers and counterfactual constructions. Even when these are introduced, 
however, the issue remains surprisingly open - so much so that it seems to me best 
to conclude that definite descriptions blur the apparently sharp logical distinction 
between particular thoughts involving a singular term and general thoughts involving 
a quantifier. 



Returning, however, to Russell himself I want finally to discuss the significance 
of the theory of descriptions within his philosophy generally. The first point is 
that Russell felt that his theory showed that there was no need to abandon his 
one-dimensional conception of meaning in favor of a Fregean theory with its all- 
encompassing distinction between Sinn and Bedeutung. Indeed, as we have seen, Russell 
felt that he could cope better than Frege with the "puzzles" of interesting identities 
and empty descriptions. This had an important implication for Russell's treatment of 
names: with no Sinn/ Bedeutung distinction Russell was committed to the view that all 
identities involving just names are trivial and that there cannot be meaningful empty 
names. So all putative counterexamples had to be handled by supposing that the names 
in question were really just descriptions under disguise. The scope of this thesis was 
massively extended by a point that Russell introduces right at the end of "On Denoting" 
(p. 427), namely that our understanding of a proposition is based upon our "acquain- 
tance" with its constituents, which are the meanings of the phrases used to express 
the proposition. For once one adds, as Russell does here, that we are not acquainted 
with matter and the minds of other people, it follows that our understanding of 
sentences that include putative names of material objects and other people cannot be 
achieved by identifying these things as the meanings of the names; instead we are 
bound to reinterpret the names as descriptions that invoke properties of things with 
which we are acquainted. 

1 shall discuss this radical doctrine of "knowledge by acquaintance" below. What 1 
want to stress here is just that it was the theory of descriptions that made this doctrine 
tenable, since it seemed to offer a way of escaping the doctrine's otherwise unaccept- 
able skeptical implications. The trick here was to suppose that "logical analysis" involv- 
ing the theory of descriptions could save the appearances of common-sense belief even 
though its obvious foundations had been removed by the doctrine of limited acquain- 
tance. This move became central to Russell's later "logical-analytic method in philoso- 
phy" (1914: v), to which I shall return below. 

Another important point is the doctrine of "incomplete symbols," and in particular 
the central claim of the theory of descriptions that phrases whose meaning at first sight 
seems to consist in denoting some object turn out, after "logical analysis," not to have 
such a meaning at all. Instead their meaning is given only "in context": in the broader 
context of the sentences in which they occur. It is striking that Frege's context princi- 
ple, which I mentioned earlier when discussing the issue of the unity of judgment, here 
enters into Russell's theory; but of course it does so only because descriptions are 
a counterexample to Russell's fundamentally non-contextual conception of meaning. 
For Russell the appeal to context is appropriate precisely where the demands of logic 
conflict with superficial grammar and its associated conception of meaning. The idea 
of such a conflict has been a deeply influential idea in the analytical tradition, giving 
rise to a very Platonist conception of "logical form" as something characteristically 
veiled by ordinary language. Wittgenstein rightly identified Russell's seminal role in 
developing this conception: 

All philosophy is a "critique of language" (though not in Mauthner's sense). It was Russell 
who performed the service of showing that the apparent logical form of a proposition need 
not be its real one. (Tractatus Logico-Philosophicus 4.003 1) 



Though by 1914 Russell took the view that "philosophy . . . becomes indistinguish- 
able from logic" (in "On Scientific Method in Philosophy") he would have repudiated 
this characterization of philosophy as just "critique of language," since for him logic 
primarily concerns the logical forms of "the various types of facts" (Papers, 8: 65); 
indeed towards the end of his life, in the 1950s, he was very critical of the way in which 
philosophers seemed primarily concerned with language. As we have already seen, and 
shall see further below, Russell's logic was always shaped by metaphysical theses (e.g. 
concerning the nature of propositions) and driven by epistemological concerns (e.g. the 
doctrine of acquaintance). Nonetheless, his own theory of descriptions, and the uses 
to which he then put it in his logical-analytic program, did seem to many philosophers 
to show that it is through the logical analysis of language that philosophers can make 
progress in resolving old debates. To a considerable extent the whole project of analyti- 
cal philosophy is founded upon this faith, and to that extent Russell's theory of descrip- 
tions remains, as Ramsey called it, "a paradigm of philosophy" (Ramsey 1931: 263n.). 

Avoiding the contradiction 

Russell's main concern in the years following the publication of The Principles of 
Mathematics was not in fact the theory of descriptions and its implications, which I have 
been discussing. Instead he was still preoccupied with his logicist project, on which he 
was now working with the Cambridge mathematician and (later) philosopher A. N. 
Whitehead, and he was, therefore, confronted by the need to avoid the perplexing 
contradiction he had discovered in 1901. 

His first thought was that the conception of an incomplete symbol he had developed 
in connection with the theory of descriptions could be put to work to show what was 
wrong with the paradoxical set R of things that are not members of themselves. Russell 
developed this thought in an ingenious way by interpreting talk of sets in terms of the 
results of substitutions within propositions and then showed that, under this interpre- 
tation, the condition of self-membership cannot be coherently expressed. Regretfully, 
however, he decided that this approach was not the whole story since it did not resolve 
paradoxes concerning propositions, such as the liar paradox, which, he felt, were so 
closely related to his own paradox that there should be a single solution for them all. 

He turned next to an idea that arose in the course of a debate with the French 
philosopher, Henri Poincare, that these paradoxes arise only because the underlying 
argument tacitly involves a "vicious circle," in that something which has been defined 
in terms of a totality is then assumed to belong to this totality. Thus Russell's set R is 
defined in terms of the set of sets which are not members of themselves, and the con- 
tradiction is then arrived at by considering whether or not R belongs to this very set. 
Similarly in the case of the liar paradox the crucial move is that whereby the statement 
made by the liar is taken to be included in the scope of the liar's own statement. So, 
Russell thought, the way to avoid all these paradoxes is to adhere to the "vicious circle" 
principle, that "Whatever involves all of a collection must not be one of the collection" 
(1910: 37). 

This principle gives rise to a hierarchy of "orders," since anything defined in terms 
of a collection of things of order n is held to be of order n + 1 and therefore not a can- 
didate for membership of the first collection. In developing the idea further, however, 



Russell went back to his earlier idea that the very definition of the paradoxical set R is 
somehow ill-formed. So he returned to the thought that our normal talk of sets involves 
incomplete symbols, and thus that sets are only "quasi-things" (1910: 81). But he no 
longer used his earlier interpretation of this talk in terms of substitutions within pro- 
positions: instead he introduced the conception of a "propositional function," a func- 
tion whose values are propositions, and proposed a way of interpreting talk of sets in 
terms of propositional functions. Under this interpretation, talk of a's membership of 
the set of things x such that fx is interpreted in terms of the truth of the proposition 
that is the value of the propositional function / A x (Russell's standard notation for 
propositional functions) for the argument a (i.e. the proposition fa). Hence the condi- 
tion of self-membership entering into the definition of R is interpreted in terms of the 
truth of the proposition that is the value of a propositional function applied to itself as 
argument. But this, Russell claims, is incoherent: he takes it that the vicious circle prin- 
ciple implies that the propositions that are the values of a propositional function should 
in no case be specified by reference to the propositional function itself, and thus that 

there must be no such thing as the value for/ A x with the argument f A x . . . That is 
to say, the symbol 'f(f A x)' must not express a proposition. (1910: 40) 

Russell reinforces this point by arguing that although there are functions of func- 
tions, in all cases functions must be of a different "type" from that of which they are 
functions; thus functions of simple individuals cannot themselves be individuals, but 
must have sufficient complexity to yield complete propositions when applied to indi- 
viduals. Similarly functions of these functions, such as the quantifiers, have to have the 
type of complexity required to yield a complete proposition in these cases. Hence, he 
says, functions cannot be arguments to themselves, for they lack the type of complex- 
ity required to yield a complete proposition in this situation. 

This line of thought generates a hierarchy of types (individuals, functions of indi- 
viduals, functions of such functions, etc.) different from the hierarchy of orders gener- 
ated by the vicious circle principle, which concerns the order of definitions. The 
resulting theory, the "ramified theory of types," is the result of merging these two hier- 
archies. There is an element of overkill in this theory, for there are now three reasons 
why the contradiction does not arise: (1) the vicious circle principle straightforwardly 
implies that R is of a higher order than its members and cannot therefore be a member 
of it; (2) Russell takes the principle to imply also that the definition underlying R, when 
spelled out in terms of propositional functions that apply to themselves, is ill-formed; 
(3) Russell also invokes a separate thesis that a function must have a different type of 
complexity from its arguments if it is to yield a complete proposition as value, which 
again implies that the definition underlying R is ill-formed. 

The main aim of the theory, however, was not to avoid the contradiction but to fulfill 
the logicist project of providing a logical foundation for pure mathematics. This was in 
a way accomplished by Russell and Whitehead in their massive, though incomplete, 
trilogy Principia Mathematica (1910-13). They had found, however, that their task was 
obstructed by the complexities of the ramified theory. For example, at a crucial point in 
the standard theory of real numbers (the least upper bound theorem), the vicious circle 
principle is violated; hence Russell and Whitehead had, in effect, to set aside this 
principle by introducing the assumption (the "axiom of reducibility") that wherever a 



prepositional function is defined in terms of a totality to which it is then required to 
belong there is a way of defining it without reference to that totality, in terms of "pre- 
dicative" functions that do not involve reference to the totality in question. As their 
critics observed, if Russell and Whitehead were going to help themselves to this assump- 
tion, then they could have simplified things a good deal by formulating the whole theory 
in terms of predicative functions in the first place. Indeed, after discussion with F. P. 
Ramsey Russell and Whitehead adopted a proposal of this kind in the second edition to 
Principia Mathematica (1927). The resulting theory, the "simple theory of types," no 
longer offers a solution to semantic paradoxes such as the liar paradox. But this is a 
positive gain since the vicious circle principle is anyway not a satisfactory resolution of 
these paradoxes, which depend on concepts such as truth whose complexities are 
separate from set theory itself. 

But there are other problems, which afflict even this modified theory. Once individ- 
uals and functions (or sets) are divided into exclusive types, there has to be a separate, 
though isomorphic, arithmetic for each type, an idea that is highly counterintuitive. 
Furthermore the validity of standard arithmetic as applied to individuals requires the 
assumption that there is an infinity of such individuals. As Russell himself recognized 
(1919: 141), this assumption, or axiom, is not logical; it is clearly metaphysical (and 
may well be false). 

To say this is to raise the question, central for the logicist project, as to what logic is. 
In Principia Mathematica logic is said to be the theory of formal inference, of inferences 
that depend merely on the logical form of the propositions involved. The distinction 
between "form" and "content" is then crucial; Russell takes it that the identification of 
the logical constants, by reference to which logical form is defined, is only a matter of 
enumeration. In his abandoned 1913 manuscript "Theory of Knowledge" he appreci- 
ates the need for some deeper theory, but remarks, "In the present chaotic state of 
knowledge concerning the primitive ideas of logic, it is impossible to pursue this topic 
further" (Papers, 7: 99). This is a surprising remark in the light of all Russell's work on 
logic. But there is no doubt that he had found the experience of coping with his con- 
tradiction a chastening experience, which had taught him that even in logic there are 
no simple answers, and thus that only "patience and modesty, here as in other sciences, 
will open the road to solid and durable progress" (Papers, 8: 73). 

To compare logic with other sciences is to invite the question whether logic 
differs from them except in respect of its subject matter of formal inference, however 
exactly that be defined. This question is particularly apposite since at least until 1911 
Russell affirmed that logic is synthetic, which might suggest that he also thought 
it is empirical. In fact, however, he held that it is a priori, and rests upon self-evident 
intuitions concerning the relationships between logical forms, which are universals 
(1912: ch. x). What is then a little odd is that he generally takes it that logical infer- 
ence is just a matter of material implication (1910: 8-9), so that although logical infer- 
ences must be in fact truth-preserving it is not required that they preserve truth in all 
possible situations. His views in this area are not easily fitted together. Later, presum- 
ably under the influence of his former student Wittgenstein, he describes logical truths 
as "tautologies," which suggests a move to a conception of them as analytic; but in 
the "Lectures on the Philosophy of Logical Atomism" (published 1918) he is still very 



Everything that is a proposition of logic has got to be in some sense or other like a tautol- 
ogy. It has got to be something that has some peculiar quality, which I do not know how 
to define, that belongs to logical propositions and not to others. (Papers, 8: 211) 

Modern set theories are based on the work of Zermelo and Von Neumann and avoid 
the problems that arise for Russell and Whitehead by sweeping away type distinctions. 
This implies that there is nothing ill-formed about the condition of self-membership, 
but Russell's paradox is avoided by denying that any well-formed condition (or propo- 
sitional function, as Russell would call it) determines a set. Indeed it is standard to have 
an axiom of "foundation," which requires that the membership of a set be founded by 
being based upon "ur"-elements which do not themselves have members. This axiom 
(proposed by Mirimanoff in 1917) implies that the condition of non-self-membership 
does not determine a set and captures the intuition which lies behind Russell's "vicious- 
circle principle," but without obstructing the development of a systematic set theory 
that can be interpreted as a foundation for mathematics. 

This does not, however, mean that modern set theory provides a vindication of 
Russell's logicist project. Russell's conception of a propositional function blurs the 
distinction between standard predicate logic and set theory, but, ironically, Russell's 
paradox itself shows the importance of maintaining a distinction here, and once this 
is drawn and set theory is provided with its own distinctive axioms concerning the exis- 
tence of sets, there is no good reason to count set theory as logic. This does not mean 
that the logicist project is altogether untenable; for one can abandon set theory in favor 
of second-order logic and try to use this to provide a foundation for mathematics. 
Whether the resulting position is satisfactory remains disputed, but even if it is, it is still 
subject to the implications of Godel's famous incompleteness theorem, which shows 
that arithmetic (and thus mathematics) cannot be completely captured within a for- 
malized theory. If logic is just the theory of formal inference, as Russell maintained, 
then Godel's theorem provides the ultimate refutation of his project of showing that 
"all mathematics is Symbolic Logic." 

Logical atomism 

When Russell was asked in 1924 to provide a personal statement of his philosophical 
position he chose to entitle it "Logical Atomism" (1956). This is a name that he began 
to use in 1914 (Papers, 8: 65) and then used in his 1918 "Lectures on the Philosophy 
of Logical Atomism." The rationale for the emphasis here on logic will be obvious; but 
what is the "atomism"? What are the "logical atoms"? 

They are atomic facts. The reference here to "facts" is due to his abandonment by 
1914 of his earlier Moorean conception of propositions. He has now adopted a form of 
the correspondence theory of truth according to which the truth of a proposition, now 
conceived of as normally a linguistic structure (though Russell also allows for imagis- 
tic mental propositions), is grounded in the perfect correspondence of logically simple 
propositions with atomic facts. I shall explain his reasons for this change of mind con- 
cerning propositions below, but since facts are said to be composed of the things that 
are the meanings of the words occurring in the proposition, it turns out that atomic 
facts differ little from old-style true elementary propositions, the propositions whose 

3 7 


truth formed the basis for the truth of propositions whose expression involves incom- 
plete symbols (see "Lectures on the Philosophy of Logical Atomism," Papers, 8: 175). 
Atomic facts are facts concerning the intrinsic qualities of, and relations between, 
particular individuals (Papers, 8: 177). But since the individuals and properties that 
constitute any fact we can talk about must be such that they are the meanings of the 
words we use, it follows that in practice the identity of atomic facts is constrained by 
the requirements of Russell's theory of meaning. And the crucial requirement here is 
that of our acquaintance with the things in question: 

A name, in the narrow logical sense of a word whose meaning is a particular, can only be 
applied to a particular with which the speaker is acquainted. (Papers, 8: 178) 

Since Russell holds that we are not acquainted with ordinary things, such as Piccadilly 
and Socrates, but only with sensory particulars which are "apt to last for a very short 
time indeed" (Papers, 8: 181), it turns out that the atomic facts we can talk and think 
about do not deal with the familiar furniture of life, but for the most part only with the 
private objects of experience. 

It is at this point that Russell's logical atomism makes contact with the doctrine of 
knowledge by acquaintance, which we encountered earlier, and thereby becomes a 
form of epistemological atomism. The core of that doctrine is expressed in his funda- 
mental epistemological principle (in "Knowledge by Acquaintance and Knowledge by 
Description") that 

every proposition which we can understand must be composed wholly of constituents with which 
we are acquainted. (Papers, 6: 154) 

With what things are we acquainted? Russell's approach is a combination of empiri- 
cism and rationalism. Through perception and introspection we gain acquaintance 
with particulars: with particular colors, noises, feelings, etc.; but through reflection 
we can also gain acquaintance with the universals of which these particulars are 
instances. This form of acquaintance involves "conception" (Papers, 6: 149-50), and is 
central to our capacity for intuitive awareness of a priori truths (1912: ch. x). 

For our present purposes, the point on which to concentrate is our acquaintance 
with particulars. Russell uses Moore's general term "sense-data" for these sensory 
particulars but emphasizes that it is for him an open question whether such 
particulars may not also exist as unperceived "sensibilia." Sense-data so conceived are 
not "in the mind," though our awareness of them is immune from error. Initially 
Russell took the view that they are subjective because their relativity to their subject's 
position and condition implies that they cannot be combined in a single public world 
(1912: ch. i); but in 1914 he switched to the view that they are physical elements 
within private spaces, which are capable of being integrated into a single public world 
unless they belong merely to dreams and hallucinations. His account of this supposed 
integration is difficult and the details cannot be discussed here, but it is for him an 
important application of his "supreme maxim in scientific philosophising" (Papers, 
8: 11) that 

Wherever possible logical constructions are to be substituted for inferred entities. 

The basic idea here is said by Russell to be similar to that involved in Ockham's razor, 
but in fact it is motivated by the need to escape the skeptical implications of his theory 
of knowledge by acquaintance. For where the apparent objects of putative knowledge 
are things of which we have no acquaintance, Russell's theory seems to imply that our 
beliefs can only ever be a matter of uncertain, speculative, inference concerning certain 
kinds of things "we know not what" whose existence is in some way implied by the exis- 
tence of the things with which we are acquainted. Russell thinks he can avoid this 
skeptical result by substituting an alternative way of thinking, which can still be rep- 
resented as a way of thinking about the apparent objects of knowledge but which, 
because it is "logically constructed" from thoughts of things (sense-data and univer- 
sals) with which we do have acquaintance, does not require speculations concerning 
the existence of any further "inferred entities." This alternative strategy shows how 
knowledge is possible while respecting the fundamental epistemological principle 
quoted earlier. 

Thus Russell takes it that knowledge of the external world is achieved on the basis 
of acquaintance with sense-data primarily by the application to these data of the logical 
principles involved in the theories of descriptions and of classes (though the integra- 
tion of private spaces into a single public one involves more than logic). The result is 
that talk and thought putatively about public physical objects is to be provisionally 
interpreted in terms of logically complex propositions about classes of sense-data; but 
because classes are themselves "logical fictions" these propositions are themselves to be 
further interpreted in terms of propositional functions. Only then can we arrive at an 
indirect specification of the atomic facts, involving only items with which we are 
acquainted, of which we are aware, and which ground our knowledge of the external 
world. But, in principle, such a specification can be arrived at; so knowledge is possible, 
or so Russell supposes. In truth the matter is more tricky than he allows because of the 
"other minds" problem. As Russell acknowledges, the construction of the external 
world involves apparent reference to the sense-data of others, for the external world 
is essentially something that transcends our own sense-data. But since we are not 
acquainted with the sense-data of others, Russell's fundamental principle implies that 
we cannot understand any propositions requiring reference to them. So we have to start 
from our own sense-data alone and build out simultaneously to other minds and the 
external world. But whether this can be done in a way that meets Russell's require- 
ments for knowledge is doubtful. 

Russell's discussion of these matters connects directly with the logical positivist 
program of the 1920s and 1930s (especially Carnap's Aufbau), though, as we shall see 
below, he himself had modified his approach in important respects by then. That 
change was an indirect result of a different problem, which caused him to substantially 
rethink his position even while he was still developing his logical atomist program. The 
background to this is his abandonment of Moorean propositions. The reason for this 
change of mind (in 1906) is that Russell ceased to find it credible that there are "objec- 
tive falsehoods," false propositions that are onto logically on a par with true proposi- 
tions, that is, facts. As we have seen, he continued to accept the existence of facts, and 



to that extent the old theory continued under a new name. But since propositions had 
been also taken to be the objects of judgment, false as well as true, Russell needed a new 
theory of judgment, which facts alone could not supply. His new theory, the "multiple- 
relation" theory, was that what had previously been conceived of as the constituents of 
a proposition that is the object of judgment should now be conceived of as terms of a 
new multiple-term relation conception of judgment; that is, instead of thinking of the 
Moorean proposition expressed by the sentence "Tom judges that A is larger than B" as 
having the logical form: 

Judges (Tom, the proposition A is larger than B) 

where this proposition is itself a complex entity somehow composed of A, B, and the 
relation of being larger than, we are to think of the same sentence (now regarded as a 
proposition itself, because it is to be the primary vehicle of truth and falsehood) as being 
such that, if true, it would correspond to a fact of the form: 

Judges* (Tom, A, B, being larger than) 

where "judges*" is the multiple-term relation that relates the subject of judgment (Tom) 
with certain objective terms (A, B, being larger than) in such a way that, together, they 
constitute a judgment that is true if and only if the objective terms constitute a 
fact - the fact that A is larger than B. 

Russell never integrated this "no proposition" theory into his logical theory. 
Although it is stated in the introduction to Principia Mathematica {PM: 43-4), its 
implications for his conception of a propositional function, which has just been defined 
as a function whose values are propositions, are not worked through, nor are its impli- 
cations for his substitutional treatment of quantifiers. Indeed it is flatly inconsistent 
with his theory of descriptions, since the complex interweaving of quantifiers and 
variables in that theory cannot be decomposed into "simple" constituents in the way 
required by the application of the multiple-relation theory to judgments involving 

Another difficulty comes from that old bugbear, the unity of judgment. For the 
theory in effect assumes that the objective terms of the multiple-term relation "judges*" 
can act as a surrogate for that which is judged, e.g. that A is larger than B. So the chal- 
lenge that the terms by themselves do not constitute a complete judgment is one that 
cannot be avoided. Russell needs to explain how an appropriate specification of the right 
truth-making fact is determined simply by the objective terms of the multiple-term rela- 
tion. One standard objection is that there is no basis for the distinction between judging 
that A is larger than B and that B is larger than A. It is arguable that this can in fact 
be handled simply by attending to the order in which the objects occur as terms of the 
relation "judges*"; but, as with descriptions, this strategy will not cope with general 
judgments involving multiple quantifiers - e.g. the judgment that all elephants are 
larger than all mice - where the bound variables obstruct the decomposition into simple 
constituents essential for Russell's approach. Wittgenstein's objection to Russell con- 
cerns a related point, that the theory permits one to judge nonsense. For unless some 
constraints are placed upon the terms of "judges*" there seems nothing to rule out a 
simple permutation of terms to generate, say, 



Judges* (Tom, being larger than, A, B) 

which would have to be a surrogate for Tom's "judgment" that being larger is A than 
B (see WITTGENSTEIN). Russell might seek to rule this out by placing type-restrictions on 
the terms of the multiple-term relation; but since type distinctions were explained in 
terms of the capacity of things to form a complete proposition Russell cannot appeal to 
them once he has embarked on his "no-proposition" perspective. 

Russell was shattered by this objection. He had spent the months of May and June 
1913 working at tremendous speed and in high spirits on a book about judgment and 
knowledge; but once he grasped Wittgenstein's point he abandoned the book (now pub- 
lished in Papers, 7) and fell almost into despair. Three years later he wrote to Ottoline 
Morrell about the crisis this event had induced: 

Do you remember that at the time when you were seeing Vittoz I wrote a lot of stuff about 
Theory of Knowledge, which Wittgenstein criticized with the greatest severity? His criti- 
cism, tho' I don't think you realized it at the time, was an event of first-rate importance in 
my life, and affected everything I have done since. I saw he was right, and I saw that I could 
not hope ever again to do fundamental work in philosophy. (1968: 5 7) 

Later philosophy 

Russell was prevented from lapsing into silence by the pressure of prior commit- 
ments at this time, most notably the Lowell lectures, which he delivered at Harvard in 
spring 1914 (published as Our Knowledge of the External World). Indeed his productiv- 
ity during 1914 is a remarkable testament to his strength of will. Once the First 
World War began he turned with some relief from the need to rethink his philosophy 
to public opposition to the war, though by 1918 he was keen to return to philosophy. 
(At the very time in early 1918 that he was standing trial for his anti-war propaganda 
and then appealing against the terms of his sentence of six months' imprisonment he 
was also delivering the lectures on the philosophy of logical atomism (Papers, 8) I have 
referred to above). 

Imprisonment, under the comfortable conditions permitted to him, turned out to 
provide the conditions of relative isolation that Russell needed to achieve a fresh start 
in his philosophy (though he also wrote his Introduction to Mathematical Philosophy, 
which is a lucid informal discussion of the main themes of Principia Mathematica). The 
starting point for this new work, which was published as The Analysis of Mind (1921), 
was William James's "neutral monism." Russell had already been thinking about this 
for some time; James's claim had been that the traditional opposition between mind and 
matter could be transcended by somehow conceiving of them as just different ways of 
thinking about something intrinsically neutral, which James called "experience." 
Russell, noting the similarities between this approach and his own account of the exter- 
nal world, develops a similar account based upon "sensations," which are like his old 
sense-data, except that he now holds that the fact that they are physical is no reason 
not to hold that they are not also mental (1995b: 143-4). The details of the construc- 
tions of the mind and of the physical world that follow are complex and unpersuasive; 
but what is nonetheless striking in the light of contemporary philosophy of mind is that 



Russell sets out a position that combines ontological monism concerning mind and 
matter with an insistence that the natural laws regulating them are not reducible either 
way (1995b: 104-5), so that the position is not a reductive monism. 

A central feature of Russell's analysis of mind is his attempt to do justice to what he 
takes to be the insights of scientific psychology, and in particular the behaviorist psy- 
chology being propounded at this time by John B. Watson. Russell cannot endorse the 
full behaviorist position; he thinks that beliefs typically involve mental imagery in a way 
that is incompatible with behaviorism. But he does endorse a broadly behaviorist 
conception of desire, and indeed refines it into a position that is recognizable as a pre- 
cursor of contemporary functionalism. Furthermore he offers a causal account of the 
content of the images that enter into beliefs and extends this into a generally causal 
account of meaning. So his analysis of mind, including mental content, is based quite 
generally upon causal considerations, and this then provides him with the materials for 
a new theory of judgment to replace that which Wittgenstein had overthrown. He does 
not, however, take full advantage of this opportunity largely because he still thinks that 
the meaning of a complete sentence is constructed from the independent meanings of 
its constituent words (1995b: 273). So it was left to Ramsey to think the matter right 
through and propose, soon after, a tentative functionalist theory of judgment. 

On the subject of knowledge, however, Russell clearly grasps the potential of his new 
causal conception of the mind. He begins The Analysis of Mind by rejecting his old con- 
ception of acquaintance, and later in the book he reinforces this break with his past by 
denying that we can obtain self-evident, certain, knowledge either by perception or 
by a priori intuition (pp. 262-6). In place of his old conception of knowledge, which, 
he now thinks, cannot rule out such "logically tenable, but uninteresting" skeptical 
hypotheses as that the world was created, with all our putative memories, five minutes 
ago (pp. 159-60), he offers "a more external and causal view" (p. 270) of knowledge. 
This is indeed precisely the view that is now familiar as "externalist"; and Russell intro- 
duces it by means of the now-familiar comparison between an accurate thermometer 
and someone with reliably true beliefs (p. 2 5 3 f f ) . 

In The Analysis of Mind Russell's presentation of this externalist conception of 
knowledge is somewhat tentative. In his last major work of philosophy, Human 
Knowledge: Its Scope andLimits (1948), Russell is much more assured and sophisticated 
in his development of this conception. His general aim here is Kantian: he seeks to 
explain how scientific knowledge is possible, but (unlike Kant) to do so within a broadly 
scientific conception of human life. The topic on which he then directs much of his 
attention is induction. This is a topic he had discussed in The Problems of Philosophy, 
where he had argued, first, that scientific knowledge is dependent upon the validity of 
the inductive principle that the greater the experience of the association of properties 
A and B the larger the probability that A and B will be found to be associated in new 
situations, and, secondly, that since this principle is presupposed in all reasoning from 
experience, it must be regarded as a self-evident a priori truth comparable to funda- 
mental truths of logic. In his later work Russell begins by arguing that this position 
is untenable, because the inductive principle is open to obvious counterexamples if 
no restrictions are placed upon the properties involved. His argument here is similar to 
that later made famous by Nelson Goodman (as "The New Riddle of Induction") (see 
GOODMAN) and he takes from it the conclusion that this is not an area within which self- 



evident truths are to be found. Instead, he proposes, it is essential to look to "scientific 
common sense" and discern the actual "postulates of scientific inference" (1948: 436). 
In the last part of the book he undertakes this task in the context of a sophisticated 
conception of knowledge as something that comes in degrees: knowledge is not just 
true belief, but there are many kinds of warrant, some merely involving reliable con- 
nections, others involving understanding and reflection, which provide the higher 
degrees of knowledge. But in the end even the higher types depend on the existence of 
the connections that justify primitive types of knowledge. Russell's conclusion is that: 

Owing to the world being such as it is, certain occurrences are sometimes, in fact, evidence 
for certain others; and owing to animals being adapted to their environment, occurrences 
which are, in fact, evidence of others tend to arouse expectation of those others. By reflect- 
ing on this process and refining it, we arrive at the canons of inductive inference. These 
canons are valid if the world has certain characteristics which we all believe it to have. 

One could not ask for a clearer statement of the externalist's justification of induction, 
though Russell is under no illusion that this will altogether satisfy the philosophical 

This late work shows Russell still capable of originality at the age of 76. These late 
writings are often neglected today. But this is a mistake. For in these late writings, he 
practices the principle he had enunciated in 1924 that "we shall be wise to build our 
philosophy upon science" (1956: 339). As a result, his writings from this period 
connect directly with contemporary debates, since from the 1970s onwards the 
"naturalization" of analytical philosophy has introduced into philosophical debate the 
requirement of harmony with scientific knowledge that Russell had recognized fifty 
years earlier. Russell is still our contemporary. 


Works by Russell 

1903: The Principles of Mathematics, London: Allen & Unwin. 

1910-13 (with A. N. Whitehead): Principia Mathematka, Cambridge: Cambridge University 

1912: The Problems of Philosophy, London: Williams & Norgate. 
1914: Our Knowledge of the External World, London: Open Court. 
1916: Principles of Social Reconstruction, London: Allen & Unwin. 
1919: Introduction to Mathematical Philosophy, London: Allen & Unwin. 
1948: Human Knowledge: Its Scope and its Limits, London: Allen & Unwin. 
1956: "Logical Atomism," in Logic and Knowledge, ed. R. C. Marsh, London: Allen & Unwin, 

pp. 323-43. (First published in Contemporary British Philosophy (first series), ed. J. H. 

Muirhead, London: Allen & Unwin, 1924, pp. 3 59-83.) 
1967: The Autobiography of Bertrand Russell 1872-1914, London: Allen & Unwin. 
1968: The Autobiography of Bertrand Russell 1914-1944, London: Allen & Unwin. 
1983-2000: The Collected Papers of Bertrand Russell, in 15 vols., London: Allen & Unwin/ 

Routledge. (The main papers cited in this chapter are listed below, and their page nos. in 

the Papers are used in text references.) 



"Knowledge by Acquaintance and Knowledge by Description," in vol. 6, pp. 148-61. (First 

published in Proceedings of the Aristotelian Society 11 (1910-11), pp. 108-28.) 
"Lectures on the Philosophy of Logical Atomism," in vol. 8, pp. 160-244. (First published 

in The Monist 28 (1918), pp. 495-527; 29 (1919), pp. 32-63, 190-222, 345-80.) 
"On Denoting," in vol. 4, pp. 414-27. (First published in Mind 14 (1905), pp. 479-93.) 
"On Scientific Method in Philosophy," in vol. 8, pp. 57-73. (Originally the Herbert Spencer 

Lecture, Oxford, November 1914.) 
"Theory of Knowledge: The 1913 MSS," vol. 7. 
"The Relation of Sense-data to Physics," in vol. 8, pp. 3-26. (First published in Scientia 16 

(1914), pp. 1-27.) 
1995a: My Philosophical Development, London: Routledge. (First published London: Allen & 

Unwin, 1955.) 
1995b: The Analysis of Mind, London: Routledge. (First published London: Allen & Unwin, 1921.) 

Works by other authors 

Godel, K. (1944) "Russell's Mathematical Logic," in The Philosophy of Bertrand Russell, ed. P. A. 

Schilpp, La Salle, IL: Open Court, pp. 123-54. 
Griffin, N. (1991) Russell's Idealist Apprenticeship, Oxford: Clarendon Press. 
Hylton, P. (1990) Russell, Idealism, and the Emergence of Analytic Philosophy, Oxford: Clarendon 

Monk, R. (1996) Bertrand Russell, vol. 1, London: Jonathan Cape. 

Pears, D. F. (1967) Bertrand Russell and the British Tradition in Philosophy, London: Fontana. 
Ramsey, F. P. ( 1 9 3 1 ) The Foundations of Mathematics and Other Logical Essays, ed. R. B. Braithwaite, 

London: Routledge and Kegan Paul. 
Watling, J. (1970) Bertrand Russell, Edinburgh: Oliver and Boyd. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

G.E.Moore (1873-1958) 


Reflecting on his long philosophical career, G. E. Moore had this to say: 

I do not think that the world or the sciences would ever have suggested to me any philo- 
sophical problems. What has suggested philosophical problems to me is things which other 
philosophers have said about the world or the sciences. 

Yet this philosophers' philosopher was lionized by the Bloomsbury literati, and his first 
book, Principia Ethica, is now included by the Modern Library Board among the one 
hundred most important nonfiction books of the century. 

Born in 1873 to a middle-class family in a London suburb, Moore went up to Trinity 
College, Cambridge, at 19. After two years studying classics, he switched to philosophy 
under the influence of his friend Bertrand Russell, but soon after that it was Moore who 
led Russell from and against idealism. With "The Refutation of Idealism" Moore set the 
direction that was to take them both to logical and philosophical analysis, and to found- 
ing, along with Ludwig Wittgenstein and the logical positivists, the philosophical move- 
ment that came to be known as "analytic philosophy." 

Moore's focus was not just on the giving of definitions or "analyses," however, 
though that certainly was central to his work, as it was for Plato. On the contrary, he 
made it clear that it is also a job for philosophy to give "a general description of the 
whole of this universe, mentioning all the most important kinds of things which we 
know to be in it." He also reflected long and deeply on what knowledge is and on how 
it might be attained. These reflections had an important impact on Wittgenstein, whose 
On Certainty is a response to Moore in epistemology (see WITTGENSTEIN). And it was 
ethics that first attracted Moore's intensely concentrated, patient attention, yielding 
that great first book of his. 

In that book, Moore introduces the analysis of concepts and properties into compo- 
nents. Consider, for example, being male and being a sibling, which come together 
through conjunction to form the complex concept of being a brother. According to 
Moore, some concepts are not thus analyzable, however, among them that of being yellow 
and that of being good (i.e. intrinsically good) , the latter of which is fundamental to ethics. 

This fundamental concept of ethics is said to be not only simple but also nonnatural, 
and irreducible to any natural property. Thus it is not analyzable in utilitarian style as 



a matter of causing or containing states of pleasure or pain. There are ways of being 
intrinsically good not dreamt of by such hedonism; being good cannot be identified with 
any one member of the plurality of intrinsic goods. Moreover, concerning any natural 
property X proposed as identical with the property of being good, there is always some 
such "open question" as: "Is having X inevitably necessary and sufficient for being 
good?" Even if we answer such a question in the affirmative, once our answer is dis- 
puted, as surely it will be if not absolutely trivial, that will show the property X and the 
property of being good not to be one and the same property. For if these were the same, 
then our question would be tantamount to the question "Is having X inevitably 
necessary and sufficient for having X?" - than which few questions can be more trivial. 
Surely this question cannot occasion the sort of controversy that always attends theo- 
ries of the good. 

Moore puts forward some excellent issues here, but his answers are not entirely sat- 
isfactory if only because based on too few distinctions (uncharacteristically so). At a 
minimum, we need the distinction between meaning analysis and philosophical, meta- 
physical analysis, as well as the distinction between concepts and properties. Perhaps 
"male sibling" can help provide a meaning analysis of "brother," so that the question 
"Is a brother a male sibling?" is indeed trivial. By contrast, "is productive of a greater 
balance of pleasure over pain than any alternative" may not give us a meaning analy- 
sis of "is the right thing to do." But it is left open that it may give a good philosophical 
analysis in any case, if one that is not obvious, not just a matter of surface meaning. 
How then are we to think of philosophical analysis compatibly with the fact that, unlike 
a kind of (surface) meaning analysis, it is far from trivial and requires reflection? This 
has proved a difficult and troubling problem for analytic philosophy, and has been not 
so much solved as shelved under the heading of "the paradox of analysis." Those who 
still care about piecemeal analysis - which still includes many philosophers, though not 
all who count themselves "analytic" - have good reason to feel nagged by this worry. 

Not through such cool analytic work did Moore attract the Bloomsburies. Most likely 
it was rather through bucking the Victorian penchant for rigid rules. These could not 
withstand Moore's probing intelligence except as rules of thumb. But Moore also dis- 
tanced himself from Bentham-Mill utilitarianism. It is certainly not just pleasure that 
fundamentally deserves our admiration and pursuit. There are things of various sorts 
that separately have that special status of the intrinsically good; prominent among these 
figure, first, the enjoyment of certain human relations and, second, the appreciation of 
things of beauty. Here Moore's milieu and upbringing may show. Talk to the starving 
millions about things of beauty, and you will be less successful than in Bloomsbury 
Still, pluralist that he is, Moore could simply accept further intrinsic goods that his list 
may have overlooked, without any major setback to his overall position. 

An act is right, Moore advises, if and only if there is no better available alternative, 
where the value of each alternative is a function not only of its own intrinsic merit but 
also of the combined value of its consequences unto eternity. These combinations are 
not just brute additions, however, but may involve special value deriving from the way 
in which they combine, as when someone is rewarded for doing well or good, or when 
someone is unrewarded or punished for doing ill or harm. 

Especially in his metaphysics and epistemology, Moore joined a tradition of common 
sense philosophy, one to which he was no doubt inherently and antecedently receptive. 



Perhaps he came to know it so deeply during his years in Edinburgh, between the end 
of his fellowship at Trinity in 1904 and the beginning of his tenure back in Cambridge 
as a lecturer in 1911. The fuller name of that tradition is, after all, Scottish Common 
Sense, which is explained mostly by the fact that its greatest early proponent 
was the Scotsman Thomas Reid. In any case, however he may have been led to this 
tradition, Moore took to it naturally, and would defend it and develop it in his own 
inimitable style. 

For Moore serious analysis required a kind of patient illumination of detail and 
nuance which is very difficult to follow attentively, and impossible for nearly anyone 
but Moore himself to produce. Indeed many of his essays in metaphysics and episte- 
mology are made up almost entirely of such minute analysis, leading up to a few bril- 
liant insights very quickly stated near the end. Take, for example, his famous "Proof of 
an External World." Except for the concluding few pages, Moore is engaged mostly in 
figuring out in great detail what "externality" could mean. That he is able to stay with 
that question through so many twists and turns, and that he does not bother to hide 
the analytic complexities illuminated by his intelligence, shows his integrity, but also 
makes him hard to read. If one stays until the end, however, the reward comes in the 
insights of the concluding pages. This is true not only of "Proof of an External World," 
but also of "Four Forms of Scepticism," wherein he is again defending his common 
sense from skeptical attack. 

That commonsensical view had already been expounded in his "Defence of Common 
Sense," wherein he describes various features of the world as he commonsensically 
believes it to be. At the center of these ontological reflections Moore inquired into 
the nature of sensory experience and its relation to physical reality in a way 
characteristically exploratory and attentive to detail. Are sense-data identical with 
physical surfaces? Are they rather nonphysical denizens of our mental world while 
representative of physical realities beyond them? Is physical reality itself to be 
viewed as somehow a construction from or analyzable or reducible to combinations of 
sense-data? Moore long struggled with such questions, but his work in this area never 
reached closure. 

Nevertheless, he felt certain enough of the core of his common sense, whatever its 
correct analysis may turn out to be, that he was willing to give it firm expression 
through a list of some of its central commitments, among them the following two: that 
he has and has for some time had a human body, which has been in contact with the 
surface of the earth, and that there have been many other three-dimensional things at 
various distances from his body. Such propositions form a first group. In a second group 
are such propositions as that he has had experiences of various sorts, and that he has 
observed various things in his surroundings at the time, and has had dreams, and other 
mental states. Finally, in a third group is the proposition that the same is true of many 
other human beings. In that paper Moore also quite explicitly claims, finally, that he 
knows with certainty propositions in his first two groups, and that the many other 
human beings of whom similar things are true also have frequently enjoyed such 
certain knowledge. 

These are of course the claims that set up his confrontation with the skeptic, where 
Moore's legendary patience and powers of analysis are very much in evidence. It is this 
work, in my judgment, that manifests a depth of insight beyond anything shown in 



Moore's other work. Not that we cannot see gaps and problems in hindsight. But his 
achievement was nonetheless real and impressive, and impressed the singularly unim- 
pressionable Wittgenstein, from whom it elicited his own best work in epistemology. In 
the next and final part of this discussion, we turn to this work of Moore's. 

Is the existence of external things just an article of faith? Certainly not, says Moore, 
and offers us a proof (which is here simplified), thus aiming to remove Kant's "scandal 
to philosophy." 

Moore's proof Here is a hand (a real, flesh and bone hand). 

Therefore, there is at least one external thing in existence. 

According to Moore, his argument meets three conditions for being a proof: first, the 
premise is different from the conclusion; second, he knows the premise to be the case; 
and, third, the conclusion follows deductively ("Proof of an External World," in 1962: 
144-5). Further conditions may be required, but he evidently thinks his proof would 
satisfy these as well. 

As Moore is well aware, many philosophers will feel he has not given "any satisfac- 
tory proof of the point in question" (1962: 147). Some, he believes, will want the 
premise itself proved. But he has not tried to prove it, and does not believe it can be 
proved. Proving that here is a hand requires proving one is awake, and this cannot be 

Does Moore adequately answer the skeptic? Many have denied it for the reason that 
he fails to rule out a crucial possibility: that our faculties are leading us astray, for 
example that we are dreaming. Aware of this objection, Moore grants, in "Certainty," 
that to know he is standing he must know he is awake ("Certainty," in 1962). The point 
"cuts both ways," however, and he would prefer to conclude that he does know he is 
awake since he does know he is standing. 

This has persuaded nearly no one. On the contrary, some have thought him com- 
mitted to an argument, M below, which is like Argument A, preceding it: 

Argument A 

Al This map is a good guide to this desert. 
A2 According to the map an oasis lies ahead. 
A3 Therefore, an oasis lies ahead. 

Argument M 

Ml My present experience is a veridical guide to reality (and I am not dreaming). 

M2 My present experience is as if I have a hand before me. 

M3 Therefore, here (before me) is a hand. 

When challenged on premise Al, our desert dullard responds: "I must know Al since 
the only way I could know A3 is through argument A, and I do know A3 ." Is this a just 
comparison? Is Moore's response to the skeptic relevantly similar? 

If Moore depends on argument M for his knowledge of M3, his response seems like 
the dullard's. The dullard is wrong to respond as he does. He must say how he knows 
his premise without presupposing that he already knows the conclusion. And Moore 
would seem comparably wrong in the analogous response to the skeptic. In explaining 
how he knows Ml, he must not presuppose that he already knows M3. 



Does Moore depend on argument M for his knowledge of M3 ? There is reason to think 
that he does not, given his emphatic acknowledgment that he cannot prove M3. After 
all, M would seem a proof of M3 just as good as Moore's own "proof of an external 
world." Moore concedes, in effect, that if he does not know that he is not dreaming then 
he does not know of the hand before him. But that is not necessarily because he takes 
himself to know M3 only through M or any other such argument. And, in any case, 
even if he is relying on some such argument, which would require making that con- 
cession, the defender of common sense has other options. 

One might, after all, make that concession only because of the following "principle 
of exclusion": 

PE If one is to know that h, then one must exclude (rule out) every possibility that 
one knows to be incompatible with one's knowing that h. 

As Moore grants explicitly, the possibility that he is just dreaming is incompatible with 
his knowing (perceptually) that he has a hand before him. And this, in combination 
with PE, is quite sufficient to explain his concession above. 

Suppose Moore is not depending on argument M for his knowledge of M3 . Although 
he recognizes his need to know he is not dreaming, suppose that is only because he 
accepts PE, our principle of exclusion. Then the sort of ridicule cast on the dullard is 
misdirected against Moore. What is more, it is not even clear that Moore must know 
how he knows he is not dreaming if he is to know M3. That is not entailed by the 
application of the principle of exclusion. All that follows from the application of that 
principle is that Moore must know that he is not dreaming, not that he must know how 
he knows this. 

In fact, however, the historical Moore did rely on something very much like argu- 
ment M (more on this below). So is he not after all exposed to the damaging compari- 
son with the desert dullard? 

Not at all. There seems no good reason why, in responding to the skeptic, Moore must 
show how he knows he is not dreaming. Of course his response to the skeptic would be 
enhanced if he could show that. But it now seems not properly subject to ridicule even if 
he is not then in a position to show how he knows he is not dreaming. The question he 
is addressing is whether he knows that he is not dreaming, and, at most, by extension, 
what grounds he might have for his answer to that question, in answering which he 
does not, nor need he, also answer the question of how he knows himself to be awake 
and not dreaming. 

It might be replied that one cannot know that here is a hand if one's belief rests 
on the unproved assumption that one is awake. According to Moore, however, things 
which cannot be proved might still be known. Besides, even though he cannot 
prove that he is awake, he has "conclusive evidence" for it. Unfortunately he cannot 
state his evidence, and the matter is left in this unsatisfactory state at the end of 
"Proof of an External World." But Moore has more to say in another paper of 
the period, "Four Forms of Scepticism" (1962: 193-223). There he takes himself to 
know for sure about the hand before him, and takes this knowledge to be based on an 
inductive or analogical argument. We are told that introspective knowledge of one's 
own sensory experience can be immediate, unlike perceptual knowledge of one's 
physical surroundings. While agreeing with Russell that one cannot know immediately 



that one sees a hand, Moore thinks, contra Russell, that he can know it for certain. 
And he disagrees with Russell more specifically in allowing knowledge for certain 
about his hand through analogical or inductive reasoning from premises known 

However, it is doubtful that any allowable form of inference - whether deductive, 
inductive, or analogical - will take us from the character of our experience to the sort 
of knowledge of our surroundings that we ordinarily claim. 

Familiar skeptical scenarios - dreaming, evil demon, brain in a vat, etc. - show that 
our experience prompts but does not logically entail its corresponding perceptual 
beliefs. Experience as if there is a fire before us does not entail that there is a fire there, 
experience as if here is a hand does not entail that here is a hand, etc. Perhaps what is 
required for one's beliefs and experiences to have certain contents entails that these 
could not possibly be entirely false or misleading. Indeed, some such conclusion follows 
from certain externalist and epistemic requirements on one's justified attribution of 
familiar contents to one's own experiences or beliefs. But even if that much is right - 
which is still controversial - one's experience or belief that here is a hand, or yonder a 
fire, might still be wildly off the mark. We cannot deduce much of our supposed knowl- 
edge of the external from unaided premises about our experience. 

As for inductive or analogical reasoning, only abductive reasoning - inference to the 
best explanation - offers much promise, but it seems questionable as a solution to 
our problem. 1 Suppose (1) that we restrict ourselves to data just about the qualitative 
character of our own sensory experience, and (2) that we view belief in a 
commonsensical external world as a theory postulated to explain the course of our 
experience. What exactly is the proposal? Is it proposed that when ordinarily we accept 
the presence of a hand before us, we do know, and know on the basis of an abductive 
inference; or is it proposed rather that in such circumstances we have resources 
that would enable us to know if only we used those resources to make effective abduc- 
tive arguments? The second, more modest, proposal is too modest, since it leaves our 
ordinary perceptual beliefs in a position like that of a theorem accepted through a guess 
or a blunder, one that we do have the resources to prove after much hard thought, but 
one that we have not come close to proving at the time when we are just guessing or 

Even the modest proposal, moreover, seems unlikely to succeed. Could we form a rich 
enough set of beliefs purely about the qualitative character of our sensory experience, 
one rich enough to permit abductive inferences yielding our commonsense view of 
external reality? This seems doubtful when we consider (1) that such pure data beliefs 
could not already presuppose the external reality to be inferred, and (2) that the pos- 
tulated commonsense "theory" of external reality must presumably meet constraints 
on abductive inference: for example that the postulated theory be empirically testable 
and also simpler and less ad hoc than alternatives (e.g. Berkeley's). These requirements 
plausibly imply that our data must go beyond detached observations, and include some 
acceptable correlations. Yet these correlations are unavailable if we restrict ourselves 
to beliefs about the character of our experience. 2 Most especially are they unavailable, 
and most especially is the postulated inference implausible, when our database is 
restricted, as it is by Moore, to introspectively known facts of one's own then present 
subjective experience, and to directly recalled facts of one's own earlier experience. (If 



deprived of the epistemic resources of testimony and of retentive memory - except 
insofar as such resources can be validated by reason-cum-introspection, which is not 
very far if at all - then there is precious little we can any longer see ourselves as 
knowing, thus deprived.) 

Accordingly, the skeptic has a powerful case against Moore's claim that our knowl- 
edge of the external is based on an inductive or analogical inference from such infor- 
mation about our experience. It is not realistic to suppose that we consciously make 
such inferences in everyday life. It is more plausible to conceive of such inferences as 
implicit or dispositional, but even this strains belief. Besides, even granted that we make 
such inferences if only implicitly, do they yield simpler and less ad hoc hypotheses than 
alternatives? That is far from clear; nor do such hypotheses seem empirically testable 
and credible simply as explanations of the purely qualitative character of our then 
present or directly recalled experience. 

Having reached a dead end, let us have some second thoughts on Moore's view of 
perceptual beliefs as inferential. Here he joined a venerable tradition along with Russell 
himself. If perceptual knowledge is thus mediate and inferential, what knowledge can 
qualify as immediate and foundational? Modern philosophy begins with Descartes's 
canonical answer to this question. 3 

Descartes had two circles, not only the big famous one involving God as guarantor 
of our faculties, but also a smaller one found in the second paragraph of his Meditation 
III, where he reasons like this: 

I am certain that I am a thinking being. Do I not therefore also know what is required for 
my being certain about anything? In this first item of knowledge there is simply a clear 
and distinct perception of what I am asserting; this would not be enough to make me 
certain of the truth of the matter if it could ever turn out that something which I perceived 
with such clarity and distinctness was false. So I now seem to be able to lay it down as a 
general rule that whatever I perceive very clearly and distinctly is true. 4 

About the cogito, I wish to highlight the inference drawn by Descartes: So I now seem 
to be able to lay it down as a general rule that whatever I perceive very clearly and distinctly 
is true. Just what is Descartes's argument in support of this general rule? Would his rea- 
soning take the following form? 

1 Datum: I know with a high degree of certainty that I think. 

2 I clearly and distinctly perceive that I think, and that is the only, or anyhow the best 
account of the source of my knowledge that I think. 

3 So my clear and distinct perception that I think is what explains why or how it is 
that I know I think. 

4 But my clear and distinct perception could not serve as a source of that knowledge 
if it were not an infallibly reliable faculty. 

5 So, finally, my clear and distinct perception must be an infallibly reliable faculty. 

The move from (1) and (2) to (3) is an inference to an explanatory account that one 
might accept for the coherence it gives to one's view of things in the domain involved. 
Descartes does elsewhere appeal to coherence at important junctures.' So he may 



be doing so here as well, although questions do arise about how Descartes views 
coherence. Does he accept the power of coherence to add justified certainty, and, in par- 
ticular, would he claim infallibility for (sufficiently comprehensive and binding) coher- 
ence as he does for clear and distinct intuition? 6 In any case, the comprehensive 
coherence of his world-view would be enhanced by an explanation of how clear and 
distinct perception comes to be so highly reliable, even infallible. And this is just what 
Descartes attempts, through his theological and other reasoning. Descartes can see that 
reason might take him to a position that is sufficiently comprehensive and interlocking 
- and thereby defensible against any foreseeable attack, no holds barred, against any 
specific doubt actually pressed or in the offing, no matter how slight. Unaided reason 
might take him to that position. Need he go any further? What is more: might one reach 
a similar position while dispensing with the trappings of Cartesian theology and even 
of Cartesian rationalism? 

Compare now how Moore might have proceeded: 

1 Datum: I know with a high degree of certainty that here is a hand. 

2 I can see and feel that here is a hand, and that is the only, or anyhow the best 
account of the source of my knowledge that here is a hand. 

3 So my perception that here is a hand is what explains why or how it is that I know 
(with certainty) that here is a hand. 

4 But my perception could not serve as a source of that degree of justified certainty 
if it were not a reliable faculty 7 

5 So, finally, my perception must be a reliable faculty. 

Moore could of course go on to say more about the nature of the perception that assures 
him about the hand. He might still say that such perception involves an implicit infer- 
ence from what is known immediately and introspectively perhaps an inductive or anal- 
ogical inference of some sort. And that might make his view more comprehensively 
coherent, but we have already seen reasons why postulating such an inference is ques- 
tionable. So we focus rather on a second alternative: Moore might well take perceiving 
to involve no inference at all, not even implicit inference, but only transfer of light, 
nerve impulses, etc., in such a way that the character of one's surroundings has a dis- 
tinctive impact on oneself and occasions corresponding and reliable beliefs. This might 
also amount eventually to a comprehensively coherent view of one's knowledge of the 
external world. And its epistemologically significant features would not distinguish it in any 
fundamental respect from the procedure followed by Descartes. 

There are other ways of opposing Moore besides that of the traditional skeptic. These 
are all based in some way or other on a key requirement of "sensitivity" for knowledge, 
one imposed on any belief candidate for knowledge, as follows: one's belief that 
p amounts to knowledge that p only if one would not believe that p if it were not the 
case that p. 

It is initially very tempting to accept the sensitivity idea common to the various forms 
of sensitivity-based opposition to Moore: namely, the skeptical, tracking, relevant- 
alternative, and contextualist approaches that share some form of commitment to that 
requirement. And, given this idea, one can then argue powerfully for the first premise 
of the skeptic's "argument from ignorance," AI, formulable by means of the following 



H I am a handless brain in a vat being fed experiences as if I were normally 

embodied and situated (see PUTNAM). 
G I now have hands. 

Here now is AI: 

(i) I do not know that not-H. 
(ii) If (i), then C (below). 
C I do not know that G. 

That lays out the skeptic's stance. Moore for his part grants the skeptic premise (ii), but 
rejects C and therefore (i). Robert Nozick's stance is different. 8 Like Moore, he rejects C. 
Like the skeptic, he affirms (i). So he must reject (ii), which he does aided by his inde- 
pendently supported account of knowledge as tracking. It is not only Nozick who rejects 
closure under known entailment; so does the relevantist, for whom in order to know 
some fact X you need not know (and often cannot know) the negation of an alterna- 
tive known to be incompatible with X, so long as it is not a "relevant" alternative. 

Having granted to the skeptic his premise A(i), contextualism is able to defend ordi- 
nary claims to know only by distinguishing the ordinary contexts in which such claims 
are made from the context where the skeptic asserts his distinctive premise in the course 
of giving argument AI. With this difference in context comes a difference in standards; 
and, because of this difference, it is incorrect to say in the skeptic's context that one 
knows G, correct though it may be to say it in an ordinary context. 

That response to the skeptic faces a problem. Moore's opponents argue that sensi- 
tivity is necessary for correct attributions of knowledge. 9 Despite its plausibility, 
however, serious objections have been published against any such requirement of sen- 
sitivity. But the problems for sensitivity do not affect a similar requirement of "safety." 
A belief is sensitive iff had it been false, S would not have held it (i.e. it would have been 
false only without S holding it), whereas a belief is safe iff S would have held it only with 
it being true. For short: S's belief B(p) is sensitive iff ~p — > ~B(p), whereas S's belief is 
safe iff B(p) — > p. These are not necessarily equivalent, since subjunctive conditionals 
do not contrapose. 10 

And now we see the problem faced by the contextualist response to the skeptic: 
namely, that an alternative explanation is equally adequate for undisputed cases (undis- 
puted, for example, between those who opt for a Moorean stance opposing the skeptic's 
distinctive premise (i) and those who opt for a contextualist stance which accepts it). 
According to this alternative explanation, it is safety that (correct attribution of) 
"knowledge" requires, a requirement violated in the ordinary cases cited, wherein the 
subject fails to know. One fails to know in those cases, it is now said, because one's belief 
is not safe. Suppose this generalizes to all uncontentious cases adduced by the con- 
textualist to favor his sensitivity requirement. Suppose in all such cases the condition 
required could just as well be safety as sensitivity. And suppose, moreover, that the prob- 
lems for sensitivity briefly noted above do not affect safety, as I have claimed. If so, then 
one cannot differentially support sensitivity as the right requirement, so as to invoke it 
in support of the skeptic's main premise. 

Here is the striking result: if we opt for safety as the right requirement then a 
Moorean stance is defensible, and we avoid skepticism. 11 That is to say, one does satisfy 



the requirement that one's belief of not-H be safe: after all, one would believe that not- 
H (that one was not so radically deceived) only if it was true (which is not to say that 
one could believe that not-H only if it was true). In the actual world, and for quite a dis- 
tance away from the actual world, up to quite remote possible worlds, our belief that 
we are not radically deceived matches the fact whether or not we are radically 
deceived. 12 

Consider, moreover, the need to explain how the skeptic's premise - that one does 
not know oneself not to be radically misled, etc. - is as plausible as it is. That require- 
ment must be balanced by an equally relevant and stringent requirement: namely, that 
one explain how that premise is as implausible as it is. 13 To many of us it just does not 
seem so uniformly plausible that one cannot ever be said correctly to know that one is 
not then being fed experiences while envatted. So the explanatory requirement is in fact 
rather more complex than might seem at first. And, given the distribution of intuitions 
here, the contextualist and the Nozickian still owe us an explanation. 

Interestingly, our distinction between sensitivity and safety may help us meet the 
more complex explanatory demand, compatibly with the Moorean stance, which I 
adopt as my own. My preferred explanation may be sketched as follows. 

Those who find the skeptic's distinctive premise plausible on the basis of the sorts of 
sensitivity considerations favored by opponents of Moorean common sense may perhaps be 
confusing sensitivity with safety, and may on that basis assess as correct affirmations 
of that premise. After all, the requirement of safety is well supported by the sorts of 
considerations adduced by Moore's opponents. Sensitivity being so similar to safety, so 
easy to confuse, it is no surprise that one would find sensitivity so plausible, enough to 
mislead one into assessing as correct affirmations of that premise. 

The plausibility of the skeptic's premise is thus explained compatibly with its 
falsity, which fits the stance of the Moorean. Of course all we really need in order to 
explain the plausibility of the skeptic's premise is that it clearly enough follows from 
something plausible enough. And the sensitivity requirement may perhaps play that 
role well enough independently of whether it is confused with a safety requirement. 
But that would still leave the question of why sensitivity is so plausible if it is just false. 
And here there might still be a role for safety to play: if this requirement of safety 
is plausible because it is true and defensible through reflection, then it may be deeply 
plausible to us simply through our ability to discern the true from the false in such a 
priori matters. Compatibly with that, some of us may be misled into accepting the 
requirement of sensitivity because it is so easy to confuse with the correct requirement, 
that of safety 14 

I have wanted to convey the power and depth of Moore's thought not only by describ- 
ing it at a lofty distance but also by engaging with it at close quarters. Despite the 
reservations I have recorded on this or that point, I hope to have made it clear how per- 
suasively right are his views on some of the most difficult and disputed issues in the 
history of our subject. But being right does not alone confer greatness in philosophy. 
From his earliest days as a thinker, Moore was not only right but also able to think for 
himself in ways opposed to the regnant orthodoxies, and to prevail as a master dialec- 
tician. One main source of his influence is now impossible to capture fully, however, 
since it resided in his viva voce contributions to the intellectual life of that golden age of 



Cambridge philosophy. This Socratic side is well conveyed by his younger Cambridge 
colleague, C. D. Broad, in an obituary for Moore: 

It was by his lectures, his discussion-classes, his constant and illuminating contributions 
to discussion at the Cambridge Moral Science Club and the Aristotelian Society, and his 
private conversation with his colleagues and pupils that he mainly produced his effects on 
the thought of his time. 


1 For Russell the "common sense hypothesis" of independent physical objects is "simpler" 
than the supposition that life is but a dream (as he explains in chapter II of The Problems of 
Philosophy). For Quine the "hypothesis of ordinary physical objects" is "posited" or "pro- 
jected" from the data provided by sensory stimulations. "Subtracting his cues from his world 
view, we get man's net contribution as the difference" (Word and Object (Cambridge, MA: 
MIT Press, 1960), p. 5). That Quine's position is deeply problematic is shown by Stroud (The 
Significance of Philosophical Skepticism (Oxford: Oxford University Press, 1984), ch. VI). 

2 This is argued by Wilfrid Sellars in "Phenomenalism," in his Science, Perception, and Reality 
(London: Routledge and Kegan Paul, 1963) (see SELLARS). 

3 The shift to discussion of Descartes may seem abrupt; however, what we find about the 
nature of immediate knowledge in that discussion has important implications for a position 
that Moore failed to explore. Skeptics who are willing to grant Descartes his immediate 
knowledge through introspection or rational intuition would need to explain exactly why 
perception could never yield such knowledge. (And what of memory?) The discussion of 
Descartes to follow is meant to highlight this issue. 

4 The Philosophical Writings of Descartes, ed. J. Cottingham, R. Stoothoff, and D. Murdoch 
(Cambridge: Cambridge University Press, 1975), vol. II, p. 24. 

5 In his Principles of Philosophy (Part TV, art. 205) for example, he notes that if we can 
interpret a long stretch of otherwise undecipherable writing by supposing that it is written 
in "one-off natural language," where the alphabet has all been switched forward by one 
letter, etc, then this is good reason for that interpretation. There he also argues for his sci- 
entific account of reality in terms of certain principles by claiming that "it would hardly 
have been possible for so many items to fall into a coherent pattern if the original principles 
had been false" (Cottingham et al., Philosophical Writings of Descartes, p. 290). Of course, if 
we join Descartes in adopting this sort of inference to an account that aids comprehensive 
coherence we will need to be able to distinguish it relevantly from the rejected abductive 
inference to an external world from introspective data about one's own experience and 
direct memories about one's past experiences. But there are important differences: for one 
thing, the present Cartesian inference is not an inference to a causal account, one with dis- 
cernible rivals that we are unable to rule out without vicious circularity. But it remains to 
be seen whether the additional theological project that Descartes next launches is or is not 
open to similar problems as those that beset the abductive inference to the external world, 
or even worse problems. We do not consider these issues which are matters of detail by 
comparison with the more abstract epistemological structure of Descartes's reasoning that 
we consider. 

6 My attribution to Descartes is tentative because of the enormous bibliography on the 
"Cartesian circle." In deference to that important tradition of scholarship, I do no more than 
suggest that there is logical space for an interpretation of Descartes that is perhaps more 



complex than many already tried, but that seems coherent and interesting. (I am myself 
convinced that this is Descartes's actual position, and defend this more fully elsewhere.) 

7 Here one would reduce Descartes's requirement of infallible certainty. 

8 Robert Nozick, Philosophical Explanations (Cambridge, MA: Harvard University Press, 1981). 

9 Keith DeRose, "Solving the Skeptical Problem," Philosophical Review 104 (1995), pp. 1-52. 

10 If water now flowed from your kitchen faucet, it would not then be the case that water so 
flowed while your main valve was closed. But the contrapositive of this true conditional is 
clearly false. 

11 I mean that we in our reflection and in our discussions in journal and seminar, avoid skep- 
ticism; we can say right here and now that we do know various things, and not just that we 
say "I know" correctly in various contexts not now our own. 

12 This sort of externalist move has been widely regarded as unacceptably circular, mistakenly, 
as I argue in detail elsewhere. 

13 When I have asked my classes to vote on that premise, generally I have found that those 
who find it false outnumber those who find it true, and quite a few prefer to suspend judg- 
ment. At every stage people spread out in some such pattern of three-way agreement- 

14 For a fuller defense of a Moorean stance in epistemology by comparison with alternative 
ideas on the epistemology marketplace, see my "How to Defeat Opposition to Moore," 
Philosophical Perspectives 13; Epistemology 13 (1999), Supplement to Nous, pp. 141-55. 


Works by Moore 

1903: Principia Ethica, Cambridge: Cambridge University Press. 
1912: Ethics, London: Williams & Norgate. 
1922: Philosophical Studies, London: Routledge and Kegan Paul. 
1953: Some Main Problems of Philosophy, London: Allen & Unwin. 
1962: Philosophical Papers, New York: Collier Books. 

Works by other authors 

Baldwin, T. (1990) G. E. Moore, London and New York: Routledge. 

Fratantaro, S. (1998) The Methodology of G. E. Moore, Aldershot and Brookfield, VT: Ashgate 

Publishing Ltd. 
Klemke, E. D. (1969) The Epistemology of G. E. Moore, Evanston, IL: Northwestern University 

Schilpp, P. A. (1968) The Philosophy of G. E. Moore, La Salle, IL: Open Court. 
Stroll, A. (1994) Moore and Wittgenstein on Certainty, Oxford: Oxford University Press. 
White, A. R. (1958) G. E. Moore, Oxford: Blackwell Publishers. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

CD. Broad (1887-1971) 


Charlie Dunbar Broad was a leading contributor to analytic philosophy of the twenti- 
eth century, known not so much for any startlingly original doctrines he propounded 
as for his formidable powers of distinction, analysis, and argument. Born in London, 
he was educated at Dulwich College and Cambridge. He entered Cambridge in 1905, 
first studying physics and chemistry in the natural science tripos and then switching 
to philosophy in the moral science tripos. The influence of Russell and Moore at 
Cambridge was then very strong and shows itself in Broad's work (see RUSSELL and 
MOORE). He published his dissertation as Perception, Physics, and Reality in 1914. For a 
period of years beginning in 1911 he served as G. F. Stout's assistant in St. Andrews, 
and in 1920 he was appointed professor at the University of Bristol, where he gave the 
course of lectures in philosophy for natural science students that became Scientific 
Thought. In 1922 he delivered the Tarner Lectures, subsequently published as The Mind 
and Its Place in Nature, and was invited to succeed McTaggart as lecturer at Cambridge. 
After McTaggart's death in 1925, he oversaw the publication of the second volume of 
McTaggart's The Nature of Existence, which served as the stimulus for writing his own 
monumental Examination of McTaggart's Philosophy. (This is the book of choice for any 
metaphysician who is sentenced to exile on a desert island.) From 19 33 until his retire- 
ment in 1953 he was Knightbridge Professor of Moral Philosophy at Cambridge. His 
other books include Five Types of Ethical Theory (1930) and two collections of papers, 
Ethics and the History of Philosophy (1952) and Religion, Philosophy, and Psychical 
Research (1953). After his death his student Casimir Lewy published his courses of 
lectures on Leibniz and Kant. 

The scope of Broad's interests was vast. Selected for attention in this article are four 
main topics: his conception of "critical philosophy," his writings on sensa and percep- 
tion, his philosophy of time, and his views on the relation of mind to matter. 

Not covered here are Broad's important contributions to the following areas: induc- 
tive logic (he sought to identify and justify some principle about the world that, if true, 
would make induction legitimate): determinism and freedom (he argued that the notion 
of "obligability" or moral responsibility is incompatible both with determinism and with 
indeterminism, making it a problematic concept); the relevance of psychic research to 
philosophy (he assessed the evidence for paranormal phenomena and identified meta- 
physical principles that would have to be given up if the reality of such phenomena 



became established); and ethics. His Five Types of Ethical Theory, a study of the ethical 
systems of Spinoza, Butler, Hume, Kant, and Sidgwick, was a widely used text; he also 
wrote influential papers in metaethics, clarifying the status of non-naturalist 
intuitionism of the sort espoused by Moore, naturalist theories of the "moral sense" 
variety, and non-cognitivist or "interjectional" theories. 

Critical versus speculative philosophy 

In a discussion of the nature and value of philosophy in the introduction to Scientific 
Thought, Broad distinguished two branches of his subject, critical philosophy and 
speculative philosophy. The first task of critical philosophy is "to take the concepts that 
we daily use in common life and science, to analyse them, and thus to determine their 
precise meanings and their mutual relations." Concepts ripe for such analysis include 
the concepts of substance, cause, place, date, duty, and self. The second task of critical 
philosophy is to test the beliefs that we constantly assume in everyday life and science, 
"resolutely and honestly exposing them to every objection that one can think of oneself 
or find in the writings of others." Beliefs subject to such critical scrutiny include the 
beliefs that we live in a world of objects that are independent of our knowledge of them 
and that every event has a cause. We may emerge from critical philosophy with ver- 
bally the same beliefs we started with, but the process will have "enabled us to replace 
a vague belief by a clear and analysed one, and a merely instinctive belief by one that 
has passed through the fire of criticism." He then went on to characterize speculative 
philosophy as follows: 

Its object is to take over the results of the various sciences, to add to them the results of 
the religious and ethical experiences of mankind, and then to reflect upon the whole. The 
hope is that, by this means, we may be able to reach some general conclusions as to the 
nature of the Universe, and as to our position and prospects in it. 

Broad noted that speculative philosophy is less certain in its results than critical phi- 
losophy, and that it must be augmented by critical philosophy if it is to be of any value. 
He engaged in both varieties of philosophy himself, but his strong suit was critical phi- 
losophy. I think it is fair to say that many analytic philosophers would cite Broad's 
definition of critical philosophy as an excellent description of what they do and Broad 
himself as an outstanding practitioner of it. 

Sense-data and perception 

Broad was one of the leading exponents of a sense-datum theory of perception. 
The term "sense-datum" was introduced by Russell and Moore; Broad himself almost 
invariably preferred the term "sensum." Though sometimes used broadly to cover 
the sensuous aspect of experience in general (however it may be analyzed), the terms 
"sense datum" and "sensum" have for Broad and other philosophers of his era a 
narrower and more precise meaning. The notion of a sensum has application only if 
one adopts an act-object analysis of sensory experience. To see what this means, con- 
sider the various types of sensory experience arranged in an order, starting with those 



of sight, passing through those of hearing, taste, and smell, and ending with bodily 
sensations like headache. At the beginning of the series, Broad claimed, it seems 
plausible to analyze a sensation of red into two components, an act of sensing and 
a red object. At the other end of the series, it does not seem plausible to analyze a 
sensation of headache into an act of sensing and a "headachy" object. Having a 
headache is not sensing something - it is sensing somehow. In the middle of the 
series (with taste and smell), it may not be obvious whether one can distinguish act 
and object. Some philosophers have assimilated the entire series to one or the other 
end of it, advocating either an act-object analysis across the board (H. H. Price) or an 
objectless "way of sensing" analysis across the board (Thomas Reid). Broad saw no 
reason to treat the entire series uniformly; he took bodily sensations to be objectless, 
but espoused an act-object analysis at least for sight, hearing, and touch. "It seems to 
me much more certain that, in a sensation of red, I can distinguish the red patch 
and the act of sensing it, than that, in a sensation of headache, I cannot distinguish a 
headachy object, and an act of sensing it" (1923: 256). The red patch that figures 
as the object-component in the sensation of red is the sort of thing Broad meant by 
a sensum. 

That is not to say that when I am seeing a ripe tomato, the tomato is a sensum. Even 
if my experience of a tomato were a total hallucination and there were no red physical 
objects in my environment, there would still be something red that I am sensing, and 
that something is a sensum. Thus sensa are not automatically to be identified with 
physical objects or even parts of them; their relation to physical things is a more 
complex affair to be discussed further below. 

The theory of sensa may be expounded further by noting some of the familiar facts 
it is meant to explain. When viewing a penny or a coffee cup tilted away from my line 
of sight, I may be certain that I am having the experience expressed by "This looks ellip- 
tical to me," even though I know that in fact the penny or the cup is not elliptical but 
round. This much is supposed to be a fact on which all parties agree. The sensum theory 
analyzes the situation as involving "the actual existence of an elliptical object, which 
stands in a certain cognitive relation to me on the one hand, and in another relation, 
yet to be determined, to the round penny" (1923: 23 7-8). This elliptical object is a 
sensum. Broad pithily conveyed the guiding motivation for positing it as follows: "If, in 
fact, nothing elliptical is before my mind, it is very hard to understand why the penny 
should seem elliptical rather than of any other shape" (p. 240). 

Generalizing from what Broad says about the penny, we may put the essential core 
of the sensum theory as follows: whenever any object x appears to a subject S to have 
a property F, it does so because S is directly aware of an item y (a sensum) that really 
does have the property F. The item y is the sensum, and its relation to x cannot in 
general be identity (since if x appears F without being F, y is F and x is not). (Certain 
restrictions are to be understood as attaching to this formulation; "appears F" is used 
phenomenally, not comparatively, and the variable F ranges over color, shape, and dis- 
tance.) It is generally held that sensa themselves, unlike physical objects, never appear 
to have any property F without really having it. This is implicit in the reason for posit- 
ing sensa in the first place: if sensa could appear to have properties they do not really 
have, we would have to posit a second tier of sensa to be the bearers of the properties 
apparently possessed by sensa in the first tier. 



What about the converse assumption, that sensa have only the properties they 
appear to have? Broad denied this, holding that sensa may be more variegated or deter- 
minate than they appear to be. This enabled him to avoid the objection that sensa would 
be indeterminate in their properties. If a sensum appears to be many-speckled without 
appearing to be exactly n-speckled for some n, Broad does not have to say that the 
sensum has speckles without having any definite number of them. 

Many writers assume that visual sensa must have only two dimensions - that they 
are extended in length and breadth, but altogether lacking in depth. Broad argued to 
the contrary that visual sense data are as fully voluminous or three-dimensional as any 
objects in physical space. Many writers assume that sensa are mental entities. This, too, 
Broad denied. According to him, they are neither mental nor physical, but have a leg 
in each realm (1925: 184). For example, they are like physical objects in having spatial 
qualities like extension and shape, but like mental things in being private to observers 
and sense modalities. Their privacy, however, does not mean that sensa are existentially 
mind-dependent; like Russell, Broad accepted it as a real possibility that there can be 
unsensed sensibilia. 

Sense-data have been out of vogue for nearly fifty years. Opposition to them has 
stemmed from two main motives. First, they are hard to accommodate within a purely 
physicalist view of the universe: if the experience of red, whether veridical or not, 
involves a literally red object, it is hard to see with what brain entities or processes this 
red object could be identified. Second, sensa make difficulties for direct realist accounts 
of perception: they are often thought to constitute a "veil" between perceivers and the 
physical world, cutting us off not only from direct perception of physical things but 
knowledge of them as well. 

What, then, are the alternatives to admitting sensa? A radical alternative is to deny 
(with Daniel Dennett and others) that there is a sensuous element in experience at all, 
in which case there would be nothing for the sensum theory to analyze. Broad would 
have dismissed this suggestion as a flagrant denial of the facts. He did, however, recog- 
nize two alternative analyses of the facts in addition to the sensum theory: the multi- 
ple relation theory and the multiple inherence theory. The first of these alternatives is 
mentioned, though not discussed, in Scientific Thought; both are discussed in The Mind 
and Its Place in Nature. 

One way to understand the differences among the three theories is to see what each 
would say about the phenomenon of perceptual relativity: the fact that the same object 
can appear to have different properties to different observers or from different view- 
points, as when water feels hot to one hand and cold to another, or a mountain looks 
blue from a distance and green close up. It would be contradictory, of course, to say 
that the same mountain is both green and blue, period. But there are three ways to state 
the facts of perceptual relativity without contradiction. First, we can say that the 
incompatible colors inhere in different subjects. This is what the sensum theory says: I 
sense one batch of sensa (blue ones) when I am viewing the mountain from afar and 
another batch of sensa (green ones) when I am standing on the summit. Second, we 
can say that the mountain looks blue as I approach it on the highway, that it looks green 
when I get there, and that on at least one of these occasions it looks to have a color that 
nothing in the situation actually has. This is what Broad called the "multiple relation 
theory of appearing"; it holds that appearing F is an unanalyzable relation between an 



object, a property, and a mind, not involving the existence of any entity that really is 
F. Third, we can say that the mountain is blue from here and that it is green from there, 
avoiding contradiction by expanding the number of places in the relation of inherence. 
This is the multiple inherence theory, in which we give up the ordinary view that the 
inherence of a color in a thing is a two-term relation between the color and the thing; 
rather, it is a three-term relation between a color, a thing, and a place or a viewpoint. 
Colors do not inhere in objects simply, but only in objects (or "regions of pervasion") 
from places (or "regions of projection"). 

The multiple relation theory has the counterintuitive consequence that objects "can 
have qualities which are different from and inconsistent with those which they seem 
on careful inspection to have" (1925: 160). The multiple inherence theory involves a 
puzzling new form of inherence; in addition, it has the puzzling consequence that the 
colors of objects are "causally adventitious" to them, in the sense that the immediate 
causal determinants of the color pervading a region lie not in that region but in some 
other region, a "region of projection" containing a suitably functioning brain. Broad 
did not think either of these theories was decisively refutable, but he found the sensum 
theory preferable on the whole. 

A further possibility is worth mentioning. One may accept the verbal formula Broad 
uses in characterizing the multiple relation theory - "an object can appear F without 
anything's being F" - without taking the relation of appearing F as unanalyzable. That, 
in effect, is what Roderick Chisholm does, analyzing "x appears F to S" as "x causes S 
to sense F-ly." He abandons the act-object analysis of sensing in favor of an adverbial 
approach, according to which to have a sensation of red is simply to sense in a certain 
way. He then analyzes the relation of an object's looking red to S as a matter of the 
object's causing S to sense redly. 

It remains to say something about Broad's views on the relation of sensing to per- 
ceiving and of sensa to physical objects. When I perceive something, I do not merely 
sense a sensum; I also believe in an object (e.g. a bell or a candle) to which the sensum 
is related. Broad devoted considerable attention to analyzing this belief and its object. 
He worked out elaborate answers to questions like these: how do physical objects cause 
sensa, and how are the places, dates, durations, shapes, and sizes of physical objects to 
be defined or known in terms of the corresponding features of sensa? The correspond- 
ing features of sensa that go by the same name are sometimes literally the same and 
sometimes not. Sensa and physical objects both have shape in the same sense, but they 
do not have location in the same sense. Sensa are literally located only in their own 
spaces (e.g. a sensum of color may be in the center of one's visual field). They may also 
be assigned locations in physical space, but only in a "Pickwickian" sense. Roughly, to 
say that a visual sensum s is "in" physical place p means this: if I turn my head to bring 
s into the center of my visual field and then follow my nose, I will bring myself closer 
and closer to p, obtaining along the way a series of sensa like s but becoming larger and 
brighter until I eventually advance beyond p and the s-like sensa disappear. 

Broad was never a phenomenalist, one who believes that physical objects are com- 
posed (or logically constructed) entirely of sensa. He believed that physical objects 
are heterogeneous composites, containing as literal parts atoms or whatever tinier 
particles are recognized by the best science of the day and containing as Pickwickian 
parts sensa belonging to the various sense realms. He also espoused something like the 



traditional distinction between primary and secondary qualities, maintaining that 
shapes inhere literally both in the scientific constituents of physical objects and in 
sensa, while colors inhere in sensa alone. His main reason for denying that colors inhere 
in physical objects was that we need to refer to the shapes of physical objects to explain 
why we sense sensa of various shapes, but do not need to assign colors to physical 
objects in order to explain why we sense colored sensa. 

Philosophy of time 

Broad had a good deal to say about the nature of space and time, including interpre- 
tations in Scientific Thought of Einstein's Special and General Theories of Relativity, 
which were then fairly new on the scene, f focus here on his more purely metaphysical 
views about time, as presented both in Scientific Thought and Examination of McTaggart's 

Some philosophers hold that only the present is real; others hold that past, present, 
and future are all equally real. Jn ST, Broad advanced a theory intermediate between 
these two, accepting the reality of the present and the past, but holding that "the future 
is simply nothing at all" (1923: 66). The time series is like a growing line, and it pos- 
sesses a direction because "fresh slices of existence" are always being added to the 
forward end of it. He drew from this the conclusion that judgments ostensibly about 
future events are neither true nor false at the times when they are made, since there is 
nothing then in existence to make them true or false (p. 73). 

Broad distinguished two aspects of time or of temporal facts, which he called the 
"extensive" (or static) and the "transitory" (or dynamic) aspects. The distinction is 
closely related to McTaggart's distinction between the A series and the B series. Call the 
relations of being earlier than, later than, and simultaneous with "B relations"; call the 
characteristics of pastness, presentness, and futurity 'A-characteristics." An A series 
is then any series of events or moments whose members have A-characteristics, and 
a B series any series whose members are related by B-relations. McTaggart noted that 
truths involving the B-relations are permanent, while truths involving the A- 
characteristics are transitory. An event that is earlier than another event is always 
earlier than it, but an event that is future will not always be future: it will become 
less and less remotely future, then it will become present, and finally it will become 
more and more past. 

A great divide in philosophies of time separates those who acknowledge the transi- 
tory aspect of time and those who reject it. Russell and many others deny it, affirming 
that temporal facts are exhausted by those involving the B-relations. Broad upheld it, 
agreeing with McTaggart that the transitory aspect of time is essential to it. He did not, 
however, believe that events become present in the way that may be suggested by 
McTaggart's language, that is, the events are already strung out and become present 
as the palings of a fence become illuminated by the passage of a spotlight. Becoming is 
not analogous to qualitative change, in which a subject that already exists acquires a 
new property; rather, to become present is just to "become," in an absolute sense. 
Broad's adherence to the transitory aspect of time is reflected instead in his insistence 
on the indispensability of tense, for tensed statements are precisely those that may 
change truth value with the passage of time. 



The indispensability of tense - the thesis that tensed verbs cannot be done away with 
in the analysis of temporal discourse - is perhaps Broad's most important thesis in his 
later philosophy of time. He opposed both Russell's analysis of tense in terms of tense- 
less copulas and B-relations and McTaggart's analysis of tense in terms of tenseless 
copulas and A-characteristics. According to Russell, an utterance of the sentence-type 
"it is now raining" means that an occurrence of rain is simultaneous with that very 
utterance; in analogous fashion, an utterance of "it has rained" or "it will rain" would 
mean that an occurrence of rain is earlier or later than that very utterance. Broad 
expressed doubt about whether the kind of self-reference involved here is really possi- 
ble and about whether tenseless verbs are anything but a philosopher's fiction. His main 
objection, however, was simply that Russell's analysis leaves out the transitory aspect 
of temporal facts. If an occurrence of rain is (tenselessly) simultaneous with a certain 
utterance, it is always simultaneous with that utterance, making any utterance of "it 
is now raining" true eternally if it is true at all. 

McTaggart's presupposition that tense is eliminable is an essential part of his noto- 
rious argument for the the unreality of time, an argument that Broad subjected to pen- 
etrating analysis. McTaggart, unlike Russell, believed that there could not be time 
without an A-series: a series of events or moments exemplifying the characteristics of 
past, present, and future. His case against the reality of time was that the A-series 
involves a contradiction: the A-characteristics are mutually incompatible, yet each item 
in any A-series must have them all. To this the obvious objection is that each event has 
all of the A-characteristics only successively and there is no contradiction in that. An 
event that is now present is not now past and future; rather, it has been future and will 
be past. But McTaggart anticipated this objection, and replied that our attempt to 
remove the contradiction only raises it anew. When we say that S has been (will be, is 
now) P, we are saying that S is P at a moment of past (future, present) time. Thus to 
say that an event has been future, is now present, and will be past implies that there is 
an A-series of moments. And this, McTaggart alleged, brings back a contradiction just 
like the original one: every moment, like every event, is past, present, and future. 

But why did McTaggart think there was a contradiction to begin with in saying that 
an event is future, present, and past, a contradiction that remains even if we add the 
qualification "successively"? To say that an event is successively future, present, and 
past is to say (if it is now present) that it was future and will be past. According to Broad, 
it is at this point in the argument that McTaggart's assumption that tense is eliminable 
plays a crucial role. Broad articulated the assumption as follows: what is meant by a 
sentence with a tensed verb or copula must be completely and more accurately expres- 
sible by a sentence in which there is no tensed verb or copula, but only temporal pred- 
icates and tenseless verbs or copulas. To highlight the fact that the more accurate 
expression must be free of tense, let us use "be" as a tenseless copula. Then McTaggart's 
claim is (e.g.) that "e was future" means "for some moment m, e be future at m & m be 
past." Well, if m be past, it is timelessly or sempiternally past. And that contradicts the 
assumption, inherent in belief in the A-series, that every moment is sometimes future 
and sometimes present. Thus Broad concludes: 

[T]he source of McTaggart's regress is that, if you take the "is" in "t is present" to be time- 
less, you will have to admit that t is also past and future in the same timeless sense of "is". 



Now this is impossible, for it is obvious that t can have these predicates only in succession. 
If, to avoid, this, you say that the "is" in "t is present" means "is now", you have not got 
rid of temporal copulas. Therefore, if you are committed at all costs to getting rid of them, 
you will not be able to rest at this stage. (1933: 314-15) 

So Broad insisted on the ineliminability of tense. Russell's attempt to eliminate tense 
in favor of the B-relations ignores the transitory aspect of time, and McTaggart's 
way of getting rid of it makes the transitory aspect contradictory. The moral Broad 
drew is that if we wish to do justice to the transitory aspect of time, we must take tense 

Broad also discussed at length the ontological categories of thing and process. One 
of the differences is that things endure literally through time, whereas a process per- 
sists only in virtue of having distinct parts or phases that exist at various moments 
within the interval. If I say "This is the same chair I sat in yesterday," I mean that lit- 
erally the same object I sat in yesterday is here now, but if I say "I am still hearing the 
same buzzing noise," I mean only that I am hearing later phases of a process whose 
earlier phases I heard before. Those who believe in the dynamic aspect of time com- 
monly hold that identity through time is a matter of thing-like endurance, while those 
who embrace a static concept of time typically hold that identity through time is really 
a matter of process-like persistence. Confounding expectations on this score, Broad 
combined his belief in the transitory aspect of time with the view that things are dis- 
pensable in ontology in favor of logical constructions out of processes. As he sometimes 
bluntly put it, a thing is just a long and boring event. 

Mind and matter 

In the concluding chapter of Mind and Its Place in Nature, Broad undertook to classify 
the various possible metaphysical theories on the relation of mind to matter, to sum up 
their strong and weak points, and to decide between them. His scheme of classification 
yielded seventeen types of possible theory, which he thought could be narrowed down 
to three or four best options and one that was most reasonable overall. I now give a 
somewhat simplified description of Broad's scheme and of his own favored alternative, 
which he called "emergent materialism." 

Suppose the X-properties of anything follow with logical or metaphysical necessity 
from some selection of its Y-properties (or the Y-properties of its parts). This could 
happen because the X-properties are identical with the Y-properties or are analyzable 
in terms of them. In this case, X-properties are reducible to Y-properties. 

Suppose the X-properties of a thing are not reducible to the Y-properties of its parts 
or the relations among them, but do follow with nomological necessity from these Y- 
properties and relations. In this case, X-properties are emergent from the Y-properties. 

With these preliminary notions granted, we can define Broad's notion of a "differ- 
entiating attribute" (or for short, simply an attribute): an attribute is a highly general 
property that is instantiated in the universe without being either reducible to or 
emergent from properties of any other type. 

Broad analyzed materiality as the conjunction of extension, publicity, persistence, 
and existential independence from observing minds. He analyzed mentality as a 



hierarchy of properties ranging from bare sentience up through the higher cognitive 
(both intuitive and discursive) and affective capacities. 

We can now arrive at most of the positions in Broad's scheme by asking the follow- 
ing questions about each of materiality and mentality: Is it instantiated in the universe 
or not? If so, is it reducible, emergent, or neither? And if it is reducible or emergent, 
with respect to what other properties is it emergent or reducible? 

If materiality and mentality are both instantiated in the universe, but neither is 
reducible to or emergent from anything else, that makes both of them attributes in 
Broad's sense, giving us the position he called "dualism." He subdivided this according 
to whether materiality and mentality can or cannot inhere in the same substance. If 
the answer is yes, we have dualism of compatibles, the position of Spinoza; if it is no, 
we have dualism of incompatibles, the position of Descartes. 

If materiality is an attribute but mentality is not, we have the family of theories 
Broad called "materialist." This subdivides according to the three ways in which men- 
tality might fail to be an attribute. If mentality is not instantiated in the universe at all, 
we have pure materialism (or what would nowadays be called eliminative materialism). 
If mentality is reducible to something else (determinates of materiality, presumably), 
we have reductive materialism. Broad discussed two chief varieties of this, "molar 
behaviorism," according to which having a mental state just means behaving in certain 
ways, and "molecular behaviorism," according to which mental processes are to be 
identified with processes in the brain and nervous system. Finally, if mentality is emer- 
gent, we have emergent materialism, according to which mental properties emerge as 
novel properties of material systems that achieve a certain degree of complexity. 

If mentality is an attribute but materiality is not, we get the family of theories Broad 
called mentalist. As with materialism, there are three possible varieties: pure mental- 
ism, reductive mentalism, and emergent mentalism. The actual mentalists Broad men- 
tions - for example Berkeley, Leibniz, and McTaggart - are all of the pure variety. It 
might be thought that phenomenalism affords an example of reductive mentalism, but 
most phenomenalists turn out to be either pure mentalists (because they hold that 
nothing in the universe exemplifies all the traits requisite for materiality) or neutralists 
(because they reduce matter to properties of "neutral" sense data in the manner to be 
described next). 

Finally, if neither materiality nor mentality qualifies as an attribute, we get the family 
of theories Broad classified as neutralist. Somewhat extravagantly, neutralism admits 
of nine subdivisions. Broad singled out two forms of neutralism as especially worthy of 
attention. First, there is the view of Samuel Alexander in Space, Time, and Deity that 
mind and matter both emerge from purely spatiotemporal attributes. Second, there is 
the view of Russell in The Analysis of Mind that materiality is not strictly instantiated 
at all (even though its various requisites are separately instantiated) and that mental- 
ity is either reducible to or emergent from properties of sense data that are themselves 
neither mental nor physical. 

Broad went on to argue that many of the seventeen types of theory can be quite 
definitely ruled out. Pure materialism and the three varieties of neutralism that say 
mentality is not instantiated can be eliminated immediately, he claimed, for mentality 
at least seems to be instantiated, and if there are seemings, there are events that instan- 
tiate mentality. He believed that both varieties of reductive materialism could also be 



ruled out. Against molar behaviorism, he pointed out that many of the mental 
states we observe within ourselves are not identical with their associated patterns 
of behavior, and he also raised the doubt whether every mental state even has a 
pattern of behavior coextensive with it. Against molecular behaviorism, he raised the 
objection that neural processes have properties (e.g. taking place swiftly or slowly) 
that do not apply to having a sensation of red. 

It remains to say something about the position Broad judged most reasonable on the 
whole, namely, emergent materialism. Because it implies that mental properties are not 
reducible to physical properties, this is a form of what is sometimes called property 
dualism, even though not a form of dualism in Broad's own sense (which requires that 
mentality be an attribute). The idea is that mental properties begin to be displayed when 
matter reaches a certain level of complexity. They are dependent on matter for their 
instantiation and are wholly determined, causally or nomically speaking, by material 
configurations. However, the mental properties of an organism "could not, even in 
theory, be deduced from the most complete knowledge of the behavior of its compo- 
nents, taken separately or in other combinations, and of their proportions and arrange- 
ments in this whole" (1925: 59). In this respect, Broad believed mental properties to be 
like the properties of chemical compounds and unlike the properties of clocks. Someone 
who had never seen a clock before could predict its behavior from the laws of physics 
together with knowledge of the clock's parts and how they are put together. By con- 
trast, someone who had never examined common salt before could not predict its prop- 
erties from the laws of physics together with complete knowledge of the properties of 
sodium and chlorine (taken separately and in other combinations) and how they are 
put together in the new compound. 

There are, of course, psychophysical laws relating mental properties to the physical 
properties of their bearers. But Broad believed these laws to be ultimate "trans-ordinal" 
laws: laws not deducible from laws already known to hold at the lower level, but dis- 
coverable instead only after we have become acquainted with objects and properties at 
the higher level. He conceded that the properties of chemical compounds, which 
he used as examples of emergent properties, might turn out with the growth of our 
physical knowledge or mathematical competence not to be emergent after all. But 
he thought that the traditional secondary qualities (whether conceived of naively as 
intrinsic properties of external things or in more sophisticated fashion as appearances 
to perceivers) were inherently emergent and that the laws connecting their instantia- 
tion with properties of microphysics would necessarily be of the trans-ordinal type. 
Not even a mathematical archangel with microscopical powers of perception, Broad 
ventured to assert, would be able to predict that ammonia smells acrid or that the sky 
looks blue. 


Works by Broad 

1914: Perception, Physics, and Reality, Cambridge: Cambridge University Press. 
1923: Scientific Thought, London: Kegan Paul, Trench, Trubner & Co. 
1925: The Mind and Its Place in Nature, London: Kegan Paul, Trench, Trubner & Co. 
1930: Five Types of Ethical Theory, London: Kegan Paul, Trench, Trubner & Co. 



1933: Examination of McTaggart's Philosophy, vol. I, Cambridge: Cambridge University Press. 
1938: Examination of McTaggart's Philosophy, vol. II, Cambridge: Cambridge University Press. 
1952a: Ethics and the History of Philosophy, London: Routledge and Kegan Paul. 
1952b: Religion, Philosophy, and Psychical Research, London: Routledge and Kegan Paul. 
1975: Leibniz: An Introduction, ed. C. Lewy, Cambridge: Cambridge University Press. 
1978: Kant: An Introduction, ed. C. Lewy, Cambridge: Cambridge University Press. 

Works by other authors 

McLaughlin, Brian (1992) "The Rise and Fall of British Emergentism," in Emergence or 
Reduction?, ed. A. Beckerman, H. Flohr, and J. Kim, Berlin: De Gruyter. 

Schilpp, P. A. (ed.) (1959) "The Philosophy of C. D. Broad," vol. X in The Library of Living 
Philosophers, New York: Tudor. (This volume contains Broad's autobiography, critical essays 
on Broad's philosophy by twenty-one authors, Broad's responses to them, and a complete 
bibliography of his writings.) 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

Ludwig Wittgenstein (1889-1951) 



Ludwig Josef Johann Wittgenstein dominates the history of twentieth-century analytic 
philosophy somewhat as Picasso dominates the history of twentieth-century art. He 
did not so much create a "school," but rather changed the philosophical landscape - 
not once, but twice. And his successors, within the broad stream of analytic philoso- 
phy, whether they followed the paths he pioneered or not, had to reorient themselves 
by reference to new landmarks consequent upon his work. He completed two diamet- 
rically opposed philosophical masterpieces, the Tractatus Logico-Philosophicus (1921) 
and the Philosophical Investigations (1953). Each gave rise to distinct phases in the 
history of the analytic movement. The Tractatus was a source of Cambridge analysis of 
the interwar years, and the main source of the logical positivism of the Vienna Circle. 
The Investigations was a primary inspiration for the form of analytic philosophy that 
flourished in the quarter of a century after the end of the Second World War, with its 
center at Oxford and its circumference everywhere in the English-speaking world and 
beyond. He taught at Cambridge from 1930 until his premature retirement in 1947. 
Many of his pupils became leading figures in the next generation of philosophers, trans- 
mitting his ideas to their students. 1 

Wittgenstein's central preoccupations at the beginning of his philosophical career 
were with the nature of thought and linguistic representation, of logic and necessity, 
and of philosophy itself. These themes continue in his later philosophy, from 1929 
onwards, although philosophy of mathematics occupied him intensively until 1944 
and philosophy of psychology increasingly dominated his thought from the late 1930s 
until his death. Having been trained as an engineer, he came to Cambridge in 1911, 
without any formal education in philosophy, to work with Russell. He was poorly read 
in the history of the subject, and intentionally remained so in later years, preferring 
not to be influenced by others. He had read Schopenhauer in his youth, and traces of 
The World as Will and Representation are detectable in the Tractatus discussion of the self 
and the will. He acknowledged the early influence upon him of the philosopher- 
scientists Boltzmann (in particular, apparently, of his Populdre Schriften) and Hertz 
(especially his introduction to The Principles of Mechanics). Apart from these figures, the 
main stimuli to his thoughts were the writings of Frege and Russell on logic and the 



foundations of mathematics. In later years, as he put it, he "manufactured his own 
oxygen." He certainly read some Kant when he was prisoner of war in Cassino, some 
of the works of Augustine, Nietzsche, Kierkegaard, and Plato, but did not cite these as 
influences upon him. 2 The only later influences he acknowledged were Oswald 
Spengler, and discussions with his friends Frank Ramsey and Piero Sraffa. 

His style of thought and writing were idiosyncratic. He was able to dig down to the 
most fundamental, and typically unnoticed, presuppositions of thought in a given 
domain. Where philosophers had presented opposing views of a topic, and debate had 
long continued polarized between alternatives, for example between idealism and 
realism in epistemology, or dualism and behaviorism in philosophy of mind, or 
Platonism and intuitionism in philosophy of mathematics, Wittgenstein did not side 
with one or another of the received options, but strove to find the agreed presupposi- 
tions common to both sides of the venerable dispute, and then challenged these. His 
insights were typically written down in highly condensed form: often a single sentence, 
a brief paragraph, or a fragment of an imaginary dialogue. Writing standard con- 
secutive prose distorted his thoughts, and, for the whole of his life, his writings were 
sequences of remarks, entered into notebooks, from which he later extracted and 
ordered the best. This, together with his great gift of simplicity of style, rich in 
metaphor, simile, and illuminating example, gives his philosophical writing power and 
fascination, as well as formidable interpretative difficulty. In one sense, he had the mind 
of an aphorist, for what is visible on the page is often no more than the trajectory of a 
thought, which the reader is required to follow through. No other philosopher in the 
history of the subject shared his cast of mind or style of thinking. The closest in spirit 
are the philosophically-minded aphorists Georg Christoph Lichtenberg (whom he much 
admired) and Joseph Joubert (with whose writings, it seems, he was not acquainted). 
During his lifetime he published only one book, the Tmctatus, and one article "Some 
Remarks on Logical Form," written for the Mind and Aristotelian Society meeting in 
1929. By the time of his death, he had more or less completed the Investigations (Part 
1), and wished it to be published. As for the rest, he left to his literary executors the 
decision on what parts of his literary remains of more than twenty thousand pages of 
notes and typescripts should be published. 3 

After the posthumous publication of the Philosophical Investigations in 1953, his 
literary executors edited numerous volumes of his unfinished typescripts and notes 
from all phases of his philosophical career. Notebooks 1914-1916 consists of prepara- 
tory notes for the Tractatus. Philosophical Remarks was written in 1929, and represents 
the stage at which the philosophy of the Tractatus was starting to crumble. Philosophical 
Grammar is an editorial compilation from typescripts written in the years 1931-4, and 
signals the transformation of Wittgenstein's thought, abandoning the philosophy of 
the Tractatus and articulating his new methods and ideas. Half of it concerns problems 
in the philosophy of mathematics, a subject which was at the center of his interests 
from 1929 until 1944. The Blue and Brown Books consists of dictations to his pupils, 
given in 1933-5. It elaborates his new philosophical methods and his transformed 
conception of philosophy, and examines problems in the philosophy of language, 
epistemology, metaphysics, and philosophy of psychology. The Remarks on the 
Foundations of Mathematics is a selection from typescripts and manuscripts written 
between 193 7 and 1944. Zettel is a collection of cuttings Wittgenstein himself made 



from typescripts written between 1929 and 1947, although most of the remarks date 
from the period 1944-7. The themes are mainly topics in the philosophy of language 
and philosophy of mind. The four volumes of Remarks on the Philosophy of Psychology 
and Last Writings on the Philosophy of Psychology are notes written between 1947 and 
1951. On Certainty and Remarks on Colour were written at the very end of his life, the 
former being unique among his works in its exclusive focus on epistemological themes. 
Apart from other minor writings, for example on Frazer's Golden Bough or aphorisms 
and general cultural observations jotted down amidst his philosophical reflections 
and gathered together in Culture and Value, five volumes of lecture notes taken by his 
students have been published. The complete Nachlass is currently being published in 
electronic form. 

Wittgenstein is unique in the history of philosophy as the progenitor of two 
profoundly opposed comprehensive philosophies. To be sure, there are continuities of 
theme between the two: the nature of linguistic representation, of logic and laws 
of thought, of the relation between thought and its linguistic expression, of the inten- 
tionality of thought and language, of metaphysics and of philosophy itself are topics 
examined in detail in the Tractatus and then re-examined in the later philosophy. There 
are also continuities of philosophical judgment. Many of the negative claims in the 
Tractatus are reaffirmed in the later works, in particular his criticisms of Frege and 
Russell, his denial that philosophy can be a cognitive discipline, his rejection of psy- 
chologism in logic and of logicism in the philosophy of mathematics. And many of the 
fundamental insights that informed the Tractatus, for example that there is an internal 
relation between a proposition and the fact that makes it true, that the propositions of 
logic are senseless but internally related to inference rules, that the logical connectives 
and quantifiers are not function names, that ordinary language is in good logical order, 
are retained in the later philosophy. Nevertheless, the insights that are thus retained 
undergo transformation, are relocated in the web of our conceptual scheme, are dif- 
ferently elucidated, and quite different consequences are derived from them. In general, 
the two philosophies represent fundamentally different philosophical methods and 
ways of viewing things. The Tractatus is inspired and driven by a single unifying vision. 
It was intended to be the culmination and closure of the great essentialist metaphysi- 
cal tradition of western philosophy. An insight into the essential nature of the elemen- 
tary proposition was held to yield a comprehensive account of the nature of logic and 
of the metaphysical form of the world, the nature and limits of thought and language. 
An ineffable metaphysics of symbolism was wedded to an equally ineffable solipsistic 
metaphysics of experience and to an atomist, realist, ontology. 

The Tractatus 

The two major thinkers whose work both inspired Wittgenstein and constituted the 
main target of his criticisms were Frege and Russell. They had revolutionized logic, 
displacing the subject/predicate logic of traditional syllogistic by the function theoretic 
logic based on the generalization of the mathematical theory of functions. Frege had 
invented the logic of generality, the predicate calculus (see FREGE). Both philoso- 
phers repudiated psychologism in logic and idealism in metaphysics and epistemology 
propounding instead forms of realism. Both had tried to demonstrate the reducibility 



of arithmetic to pure logic, Frege in The Basic Laws of Arithmetic (1893, 1903) and 
Russell, together with Whitehead, in Principia Mathematica (1910). It was, above all, 
their conception of logic that set the agenda for the young Wittgenstein. 

Frege and Russell thought that logic was a science with a subject matter. The propo- 
sitions of logic, they held, are characterized by their absolute generality. On Frege's view 
they are perfectly general propositions concerning sempiternal relations between 
thoughts (propositions), articulating laws of truth valid for all thinking. According to 
Russell, logic is the science of the perfectly general. Its propositions are descriptions of 
the most general facts in the universe. Hence neither would have considered a simple 
tautology such as "Either it is raining or it is not raining" as a proposition of logic, but 
would have conceived of it as an instantiation of the logical proposition "(p) (p v ~p)." 
Both tended to view rules of inference ("laws of thinking") as related to the proposi- 
tions of logic ("laws of truth") somewhat as technical norms specifying a means to an 
end are related to laws or regularities of nature. The laws of truth according to Frege 
describe the immutable relations between thoughts (propositions) irrespective of their 
subject matter; according to Russell, they are the most general laws governing the facts 
of which the universe consists. Accordingly, rules of inference are technical norms, 
dependent on such general laws, ensuring that if one wishes to think correctly, i.e. infer 
only truths from truths, one will do so. The logical systems the two philosophers had 
invented were axiomatized, and they viewed the axioms as self-evident truths. Frege 
conceived of thoughts and of the two truth-values as logical objects, and of the notions 
of object, concept, first- and second-level function as ultimate summa genera, drawing 
ontological distinctions "founded deep in the nature of things." The logical connectives 
he thought to be names of logical entities, unary or binary first-level functions mapping 
truth-values on to truth-values, and the quantifiers to be names of second-level func- 
tions. Russell held that terms such as "particular," "universal," "relation," "dual 
complex," are names of logical objects or "logical constants" signifying the pure forms 
which are the summa genera of logic, the residue from a process of generalization which 
has been carried out to its utmost limits. We understand such expressions, he thought, 
on the basis of "logical experience" or intuition. Both philosophers held natural lan- 
guage to be logically imperfect, containing vague and ambiguous expressions or names 
without reference, and hence, Frege thought, allowing the formation of sentences 
without a truth-value. They viewed their own notations as logically perfect languages. 
From the post-Wittgensteinian perspective, Frege and Russell were radically mistaken 
about the nature of logical truths (conceiving of them as essentially general), about the 
nature of logical necessity, about the content of logical truths, about the status of the 
axioms of logic, about the character of the logical connectives and quantifiers, and 
about the relation between the truths of logic and rules of inference. If we are any 
clearer on these matters than they, it is largely due to Wittgenstein. 

In the Tractatus, Wittgenstein accepted some of the salient doctrines of Russell and 
Frege. Like them, he adopted a (different) variant of metaphysical realism in the 
Tractatus ontology of simple sempiternal objects, of complexes, and of facts. He 
accepted unreflectively the assumption that the fundamental role of words is to name 
entities (although this role was denied to logical operators and to categorial expressions) 
and of sentences to describe how things are in reality. He thought that there must be a 
connection of meaning between words and the entities they name, that language 



acquires content by means of such a connection with reality. He agreed with their 
antipsychologism in logic. He accepted Frege's demand of determinacy of sense, 
although unlike Frege, he thought that the vagueness of natural language was merely 
superficial and analyzable into disjunctions of determinate possibilities. And, like Frege, 
Russell and many others, he assumed that the logical connectives and quantifiers 
are topic-neutral. Some of these commitments he was later to abandon, others he 

Unlike Frege and Russell, Wittgenstein held that ordinary language is in good logical 
order. For logic is a condition of sense, and insofar as sentences of ordinary language 
express a sense, convey thoughts, they are in good order - any appearance to the con- 
trary (e.g. vagueness) being a feature of the surface grammar of expressions, which 
will disappear on analysis. Insofar as they fail to express a sense, they are ill-formed 
pseudo-sentences. Hence it is not the task of philosophy to devise a logically ideal lan- 
guage, although devising a logically perspicuous notation will enable the philosopher 
to lay bare the true logical forms of thoughts, which are obscured by the surface 
grammar of ordinary language. According to the Tractatus the fundamental function 
of language is to communicate thoughts by giving them expression in perceptible form. 
The role of propositions (sentences with a sense) is to describe states of affairs, which 
may or may not obtain. If the state of affairs depicted by a proposition obtains, then 
the proposition is true, otherwise it is false. Propositions are composed of expressions. 
Logical expressions apart, the constituent expressions in a proposition are either ana- 
lyzable, definable by analytic definition or paraphrase, or unanalyzable. Unanalyzable 
expressions are simple names, which are representatives of simple objects. The simple 
objects are the meanings of the names. Hence names link language to reality, pinning 
the network of language on to the world. Names have a meaning only when used as 
representatives, and they are so used only in the context of a proposition. The elemen- 
tary (logically independent) proposition is a concatenation of names in accordance 
with logical syntax. It does not name anything, pace Frege (who thought sentences 
name truth-values) and Russell (who thought they name complexes), but depicts a (pos- 
sible) state of affairs, which is isomorphic to it given the rules of projection, and asserts 
its existence. The names in an elementary proposition must possess the same combi- 
natorial possibilities in logical syntax as the metaphysical combinatorial possibilities of 
the objects in reality that are the constituents of the state of affairs represented. The 
logical syntax that underlies any possible means of representation mirrors the logico- 
metaphysical forms of reality. Pace Frege and Russell, the assertion sign has no logical 
significance. Unlike Frege, who thought that there were alternative analyses of propo- 
sitions, and unlike Carnap, who, in the 1930s, thought that we can choose between 
different logics, Wittgenstein thought that analysis is unique and that in logic there are 
no options. 

The metaphysics of the Tractatus was realist (as opposed to nominalist), pluralist (as 
opposed to monist), and atomist. The sempiternal objects that constitute the substance 
of all possible worlds include properties and relations of categorially distinct types. It is 
far from clear what kinds of things Wittgenstein had in mind, but they are arguably 
such items as minimally discriminable shades of color, tones, etc. as well as spatio- 
temporal points in the visual field. Objects are simple (this is mirrored by the logical sim- 
plicity, i.e. unanalyzability, of their names). They have internal and external properties. 



Their internal properties constitute their (essential) form: their combinatorial possibil- 
ities with other objects (this is mirrored by the logico-syntactical combinatorial pos- 
sibilities of their names). Different objects belonging to the same ontological category 
(e.g. different shades of color) have a common form (namely, color). The external prop- 
erties of objects are accidental: their contingent concatenations with other objects to 
form actual states of affairs. A state of affairs is a possible combination of objects (e.g. 
that such and such a spatiotemporal point is a certain shade of such and such a color). 
The obtaining or non-obtaining of a state of affairs is a fact (hence there are positive 
and negative facts). Elementary states of affairs are "atomic" or "independent," that is, 
each such state of affairs may obtain or not obtain while all other elementary states of 
affairs that obtain remain the same. This is reflected by the logical independence of the 
elementary proposition, which has no entailments. The metaphysics of experience in 
the Tractatus was apparently a form of empirical realism and transcendental solipsism 
(cf. Kant's empirical realism and transcendental idealism). The empirical self that is 
studied by psychology is not an object encountered in experience, but a (Humean) col- 
lection of experiences. The metaphysical self, which is the concern of philosophy, is a 
limit of experience. It is the willing self, the bearer of good and evil. 

Sentences are expressions of thoughts. But thought itself is a kind of language, 
composed of thought-constituents. The form of a thought must mirror the form of 
reality no less than a proposition. Natural language is necessary for the communica- 
tion of thoughts but not, it seems, for thinking - which can be effected in the "language 
of thought." It is mental processes of thinking and meaning that inject content into 
the bare logico-syntactical forms of language. What pins a name on to an object in 
reality that is its meaning (Bedeutung) is an act of meaning (meinen) by the name of that 
object. What renders a licit concatenation of signs a living expression of a thought is 
the employment of the method of projection, which is thinking the sense of the 
sentence, i.e. meaning by the sentence such and such a state of affairs. Hence the 
intentionality of signs is derived from the (intrinsic) intentionality of thinking and 
meaning (meinen). 

The Tractatus account of the intentionality of thought and language is informed by 
the insight that thought and proposition alike are internally related to the fact that 
makes them true. The thought or proposition thatp would not be the thought or propo- 
sition it is were it not made true by the fact that p and made false by the fact that not 
p. What one thinks when one thinks truly thatp is precisely what is the case, and not 
something else (such as a Fregean Gedanke), which stands in some relation to what is 
actually the case. But what one thinks when one thinks falsely thatp is not what is the 
case (since what one thinks does not obtain). Yet one does not think nothing. Indeed, 
what one thinks is the same, no matter whether one thinks truly or falsely. The picture 
theory of thought and proposition provided a logico-metaphysical explanation of how 
it is possible to satisfy the demands consequent upon these internal relations. It 
attempts to explain how it is possible for a thought to determine what state of affairs 
in reality will make it true, how it is possible for the content of a thought to be precisely 
what is the case if it is true and yet to have a content even if it is false, how it is 
possible that one can read off from a thought, in advance of the facts, what will make 
it true, and how it is possible for the "mere signs" of language to be intentional, i.e. for 
a name to reach up to the very object itself of which the name is the name and for the 



sentence to describe the very state of affairs the existence of which will make true the 
proposition expressed. 

Every representation is a picture of a possibility. A proposition or thought is a logical 
picture, whose simple constituents name sempiternal objects with determinate form. 
There is a metaphysical harmony between language and thought on the one hand and 
reality on the other; for when one thinks truly that p, what is the case is that p\ and 
when one thinks falsely that p, what one thinks is precisely what is not the case. This 
"pre-established harmony" is orchestrated by a metaphysics of symbolism. Only simple 
names can represent simple objects. Simple names have a meaning but no sense. 
Relations too are objects, and only relations can represent relations; hence in the propo- 
sition "aRb," it is not "R" that represents the relation that a stands in to b, but rather it 
is that "R" stands to the right of "a" and to the left of "b" (in this notation). Only facts 
can represent facts, and sentences - in their symbolizing capacity - are facts, which are 
used to describe how things are. For it is the fact that the constituent names are 
arranged as they are (in accordance with logical syntax) that says that things are thus 
and so. Sentences have a sense but no meaning. 

The possible states of affairs in reality are determined by the language-independent 
combinatorial possibilities of objects. Every elementary proposition depicts a possible 
state of affairs. It is true if the possibility depicted obtains, false if it does not. It is of the 
essence of the proposition with a sense to be bipolar, i.e. to be capable of being true and 
capable of being false. 4 This mirrors the metaphysical truth that it is of the nature of 
states of affairs that they either obtain or fail to obtain. The sense of a proposition is its 
agreement and disagreement with the existence and non-existence of states of affairs. 
For the proposition that p agrees with the fact that p and disagrees with the fact that 
not-p. What one thinks when one thinks that p is a possibility, a possibility which is 
actualized if one's thought is true and is not if one's thought is false. Hence one can 
read off a proposition or thought (which is a kind of proposition) what must be the case 
for it to be true, and what one thinks when one thinks that p is precisely what is the 
case if one's thought is true and what is not the case if one's thought is false, and is the 
very same thought no matter whether it is true or false. 

The logical connectives are not names of functions, but rather signify truth- 
functional operations on propositions. The quantifiers are construed as operators upon 
a propositional function (e.g. "fx") which is a logical prototype collecting all proposi- 
tions of a certain form (whose values are all those propositions obtained by substitut- 
ing a name for the variable), hence generating logical sums or products of such sets of 
propositions. All possible molecular propositions can be generated by truth-functional 
operations upon elementary propositions. Hence all logical relations are determined 
by truth-functional combinations of propositions. A molecular proposition p entails 
another proposition q if and only if the sense of q is contained in the sense of p, i.e. if 
the truth-grounds of p contain the truth-grounds of q. The various operators are inter- 
definable, and reducible to the single operation of joint negation, namely "not . . . and 
not ..." Among the truth-functional combinatorial possibilities of a given number of 
elementary propositions, there will always be two limiting cases (1) in which the propo- 
sitions are so conjoined as to be true irrespective of the truth-values of the constituent 
propositions and (2) false irrespective of their truth- values. The former is a tautology 
and the latter a contradiction. These are the propositions of logic. Since they are, respec- 



tively, true and false irrespective of how things are, they are wholly without any 
content, and say nothing about how things are in reality. So by contrast with other mo- 
lecular propositions which are true under certain conditions (i.e. for certain assign- 
ments of truth-values to their constituents) and false under others, the propositions 
of logic are unconditionally true or false. Hence they are said to be senseless, to have, 
as it were, zero-sense. All tautologies say the same thing, namely nothing. But differ- 
ent tautologies may nevertheless differ, for every tautology is a form of a proof 
(since every tautology can be rewritten in the form of a modus ponens), and different 
tautologies reveal different forms of proof. It is a mark of the propositions of logic, 
Wittgenstein held, that in a suitable notation they can be recognized as such from the 
symbol alone. He invented a special notation to display this, his T/F notation. Instead 
of writing molecular propositions by means of symbols for logical connectives, he 
used truth-tables as propositional signs. Here it is immediately perspicuous from the 
sign alone whether a proposition is a tautology, and if so, it is visibly evident that 
it cannot be false. It is equally evident whether one proposition follows from another, 
i.e. whether the truth-grounds of one contain those of another. This showed, he 
thought, the nature of the propositions of logic and their categorial difference from 
empirical propositions. 

This conception of logical truth made clear how misleading was the Frege/Russell 
axiomatization of logic, with its appeal to self-evidence for the axioms. Their axioms 
were not privileged by their self-evidence. They were tautologies no less than their 
theorems. They were not "essentially primitive," nor were Frege's and Russell's theo- 
rems essentially derived propositions, for "all the propositions of logic are of equal 
status," namely tautologies that say nothing. Hence too, contrary to Frege and Russell, 
the propositions of logic have no sense, and describe nothing. In an important sense, 
the propositions of logic have no subject matter, and logic is misconstrued as the science 
of the most general laws of truth or of the most general facts in the universe. 
Consequently, the propositions of logic do not constitute the foundations for the 
elaboration of technical norms of thinking on the model of the relation between laws 
of nature and technical norms for achieving desired ends. Rather, every tautology is 
internally (not instrumentally) related to a rule of inference or form of proof. 

The conception of logic in the Tractatus was still flawed. But its flaws, which 
Wittgenstein was later to expose, did not significantly affect the criticisms of the 
Fregean and Russellian conceptions of logic. According to the Tractatus the only 
(effable) necessity is logical necessity. Every well-formed proposition with a sense 
must be bipolar. What philosophers had hitherto conceived of as categorial (or 
formal) concepts, such as object, property, relation, fact, proposition, color, number, etc. 
are, Wittgenstein argued, expressions for forms, which are represented by variables, 
rather than by names. Hence they cannot occur in a fully analyzed proposition with a 
sense. One cannot say that, for example, one is a number, that red is a color, or that 
A is an object, for such pseudo-propositions employ a formal concept as if it were a 
genuine concept, and they are not bipolar. Hence such metaphysical pronouncements 
(which attempt to describe non-logical necessities) are nonsense - ill-formed 
conjunctions of signs. But what such pseudo-propositions try to say is actually shown 
by genuine propositions which contain number words, color names, or other names 
of objects. It is shown by features of the expressions in such propositions, namely by 



the forms of the expressions - their essential combinatorial possibilities. These are 
represented by the variable of which the meaningful names are substitution-instances. 
An immediate consequence of this is that most of the propositions of the 
Tractatus which delineate the necessary forms of language and reality are nonsense. 
Hence Wittgenstein's penultimate remark in the book: "My propositions serve as 
elucidations in the following way: anyone who understands me eventually recog- 
nizes them as nonsensical, when he has used them - as steps - to climb up beyond 

Hence too, the conception of philosophy advocated for the future is not the practice 
exhibited in the book. The Tractatus consists largely of sentences that are neither bipolar 
propositions nor tautologies. They attempt to describe the essence of the world, of lan- 
guage, and of logic, and of the essential relations between them. But this is an attempt 
to say the very things that cannot be said in language, but are rather shown by lan- 
guage. What is thus shown is indeed ineffable. Hence metaphysics, the attempt to dis- 
close the essential natures of things, is impossible. Once the correct logical point of view 
has been achieved, once the world is seen aright, the task of the Tractatus is completed. 
The task of future philosophy is analysis: clarification of philosophically problematic 
propositions which will elucidate their logical forms or clarify why and where (in the 
case of putative metaphysical propositions) they fail to accord with the rules of logical 
grammar. Future philosophy will not be a theory, nor will it propound doctrines or 
attain knowledge. It will be an activity of logical clarification. Philosophy, thus con- 
ceived, is a critique of language. 

The role of the Tractatus in the history of 
analytic philosophy 

In six respects the Tractatus introduced the "linguistic turn" in philosophy. First, it set 
the limits of thought by setting the limits of language: by elucidating the boundaries 
between sense and nonsense. This put language, its forms and structures, at the center 
of philosophical investigation. Second, the positive task for future philosophy was the 
logico-linguistic analysis of sentences. The logical clarification of thoughts is to proceed 
by the clarification of propositions - sentences with a sense. Third, the negative task of 
future philosophy was to demonstrate the illegitimacy of metaphysical assertions by 
clarifying the ways in which attempts to say what is shown by language transgresses the 
bounds of sense. Fourth, the Tractatus attempted to clarify the essential nature of the 
propositional sign by elucidating the general propositional form, that is, by giving "a 
description of the propositions of any sign-language whatsoever in such a way that every 
possible sense can be expressed by a symbol satisfying the description, and every symbol 
satisfying the description can express a sense, provided that the meanings of names are 
suitably chosen." Fifth, the logical investigation of phenomena, the unfolding of their 
logical forms, which was not undertaken in the book, is to be effected by logical analy- 
sis of the linguistic descriptions of the phenomena. (The first moves in carrying out 
this task were taken in the 1929 paper "Some Remarks on Logical Form," whereupon 
the whole project collapsed.) For the logical syntax of language is and must be 
isomorphic with the logico-metaphysical forms of the world. Sixth, the greatest 
achievement of the book, as seen by the Vienna Circle, was its elucidation of the nature 



of logical necessity. This was patently made by an investigation of symbolism. That one 
can recognize the truth of a logical proposition from the symbol alone was held to 
contain in itself the whole philosophy of logic. 

Many of these claims were later to be repudiated. But they heralded the linguistic 
turn, which was executed by the Vienna Circle, and, in a different way, by Wittgenstein 
himself in his later philosophy, and by Oxford analytic philosophy. The Tractatus was a 
paradigm of analytic philosophy in its heroic or classic phase in the interwar years. It 
was the major inspiration of Cambridge analysis and of logical positivism. Its program, 
as understood both in Cambridge and in Vienna, committed one to the method of 
logico-linguistic analysis of complex expressions into their simple unanalyzable con- 
stituents. It encouraged the program of reductive analysis and its mirror image, logical 
construction. It cleaved to the thesis of extensionality holding all non-extensional con- 
texts to be either eliminable, merely apparent, or illicit. It repudiated the intelligibility 
of putatively synthetic a priori propositions, insisted that the only necessity is logical 
necessity and denied any sense to the propositions of logic. Hence it seemed to provide 
the foundations for what the Vienna Circle hailed triumphantly as "consistent empiri- 
cism," for it denied that pure reason alone can attain any knowledge of the world. It 
held metaphysics to be nonsense (the Circle averted their gaze from, or quickly con- 
demned and passed over (Neurath), or attempted to circumvent (Carnap), its para- 
doxical ineffability claims). And it allocated to philosophy a sui generis analytic role and 
a status wholly distinct from that of science. Schlick, the leading figure in the Circle, 
went so far as to characterize the Tractatus as "the turning point in philosophy," the 
deepest insight into what the task and status of philosophy should be. 

Wittgenstein's influence upon the Vienna Circle was second to none. Indeed, the 
principle of verification itself was derived from conversations with Wittgenstein in 
1929/30, and read back into the Tractatus. Members of the Circle spent two academic 
years reading through the book line by line, abandoning some of its claims and accept- 
ing others. They abandoned the picture theory of the proposition, the doctrine of 
showing and saying, and most of the ontology of logical atomism. But what they 
accepted was crucial: the account of the nature and limits of philosophy, the concep- 
tion of logic and logical necessity, and the program of the logical analysis of language 
(see AYER, CARNAP, HEMPEL, and OUINE). These ideas, interpreted and sometimes seriously 
misinterpreted, were pivotal to their work. The most important misinterpretation con- 
cerned the Tractatus account of logic. Members of the Circle agreed with the criticisms 
of the Fregean and Russellian misconceptions of the nature of logic, and welcomed the 
view that the propositions of logic are vacuous (senseless). But they gave a conven- 
tionalist interpretation to Wittgenstein's account of logic which was far removed from 
his conception. They thought of the logical connectives as arbitrary symbols introduced 
to form molecular propositions, whereas Wittgenstein had argued that they are essen- 
tially given by the mere idea of an elementary proposition. Where he viewed the truths 
of logic as flowing from the essential bipolarity of the proposition, they conceived of 
them as following from the truth-tabular definitions of the logical connectives - hence 
as true in virtue of the meanings of the logical operators. A logical truth therefore was 
held to be the logical consequence of conventions (definitions). Wittgenstein, by con- 
trast, had argued that the senseless truths of logic reflect the logical structure of 
the world. Logic, far from being determined by convention, is transcendental. In the 



1930s, when he turned to reconsider his earlier conception, Wittgenstein not only 
reformulated his views but also vehemently criticized the conventionalism of the Circle. 
Far from following from the meanings of the logical connectives, the truth of the propo- 
sitions of logic, he argued, is constitutive of their meanings. 

The collapse of the Tractatus vision 

Already in the Tractatus Wittgenstein had taken note of the fact that determinates of a 
determinable, e.g. red and green, are mutually exclusive: if A is red all over, it follows 
that it is not green (or blue or yellow, etc.) all over. At the time, he thought that this 
showed that "A is red" is not an elementary proposition, and that its entailments would, 
on analysis, be clarified as following from its truth-functional composition out of ele- 
mentary propositions. When he returned to philosophy after a hiatus of a decade, he 
realized that this was misconceived. There are irreducible logical relations of exclusion 
or implication which are determined not by truth-functional composition, but by the 
inner structure of elementary propositions. He tried to budget for this by abandoning 
the topic neutrality of the logical connectives and drawing up truth-tables specific to 
the "propositional system" (i.e. the system of determinates of a determinable) to which 
a given elementary proposition belongs. In the case of color, the conjunction of "A is 
red all over" and "A is green all over" is nonsense. Hence the truth-value assignment 
"TT" must be excluded from such conjunctions by a special rule of syntax. But this con- 
cession, he rapidly realized, spells the death-knell for the philosophy of logical atomism, 
and strikes at the heart of the Tractatus. For the independence of the elementary propo- 
sition was the pivot upon which turned the whole conception of logic and the inef- 
fable metaphysics of the book. Without it, the idea that the logic of propositions depends 
only upon the bipolarity of the elementary proposition collapses. The significance of 
the T/F notation as revealing the essential nature of logical propositions and relations 
evaporates, precisely because there are logical relations that depend upon the inner 
structure of elementary propositions. Since the logical operators are not topic neutral, 
separate truth-tables would have to be drawn up for each propositional system. The idea 
that there is a general propositional form, according to which every proposition is a 
result of successive applications to elementary propositions of the operation of joint 
negation must likewise be relinquished. So too must the thought that generality can be 
analyzed into logical sums and products, and that the quantifiers can be given a 
uniform topic-neutral analysis. 

As the logical theory of the Tractatus collapsed, so too did the metaphysics. It was 
wrong to say that the world consists of facts rather than of things. Rather, a descrip- 
tion of the world consists of statements of facts, not of an enumeration of things. But 
the statement of a fact just is a true statement. One cannot point at, but only point out, 
a fact. And to point out a fact just is to point out that things are thus and so, that is, to 
make a true assertion. Facts are not concatenations of objects. Unlike concatenations 
of objects, and unlike states of affairs, facts have no spatiotemporal location. The fact 
that a circle is red is not composed of redness and circularity concatenated together, 
since facts are not composed of anything and do not have "constituents." The proposi- 
tion thatp is only "made true" by the fact thatp in the sense in which being a bache- 
lor makes one unmarried. All it means is that the proposition thatp is true if, in fact, 



things are as it says they are. The conception of absolutely simple sempiternal objects 
was incoherent. For the notions of simplicity and complexity are relative, not absolute. 
To call spatiotemporal points, properties, or relations "objects" is a misuse of language. 
What had appeared to be objects that had to exist are in fact samples which we employ 
in explaining the meanings of certain ostensively defined expressions in the language. 
As such, they belong to the means of representation, not (like the postulated "objects" 
of the Tractatus) to what is represented. 

As the metaphysics collapsed, so too did the picture theory, the conception of iso- 
morphism between language and reality, and the account of intentionality. What had 
seemed like an internal relation between the proposition thatp and the fact thatp which 
makes it true was no more than the shadow cast upon reality by an intra-grammatical 
relation between the expressions "the proposition that p" and "the proposition made 
true by the fact that p." There is an internal relation here, but it is forged in language - 
in the grammatical rule that permits the inter-substitution of these expressions - not 
between language and reality. Hence it was mistaken to think that reality must have a 
certain metaphysical form which must be reflected in the logico-syn tactical forms of lan- 
guage in order for this internal relation to obtain. The intentionality of thought and 
proposition, which had seemed to demand a pre-established metaphysical harmony 
between language and reality, is fully explained by reference to intra-grammatical con- 
nections between expressions. The thought or expectation that it will be the case thatp 
does not "anticipate reality"; rather, only what satisfies the description "it is the case that 
p" will be called "the fulfillment of the expectation that it will be the case that p." Of 
course one can "read off" from the thought what will make it true, since the expression 
of the thought contains the description of the state of affairs the obtaining of which is 
called "the confirmation of the thought." Of course what one thinks, when one thinks 
thatp, is what is the case when one's thought is true, but this is not a strange form of 
identity or coincidence between a shadowy possibility and an actuality. Rather the ques- 
tion "What is being thought?" and "What is the case?" here receive the same answer. 

The metaphysics of symbolism of the Tractatus was in fact a mythology of symbol- 
ism. The meaning of a name is not an object of any kind. What is legitimate about the 
role which the Tractatus simple object was invoked to fulfill is in fact played by defining 
samples used in ostensive definitions, e.g. of color words. But the sample pointed at in 
the ostensive definition "This is black" is part of the means of representation, to be used 
as an object of comparison and standard of correct application of the word "black." 
Names derive their meanings not from objects in the world which they represent, but 
from explanations of meaning, of which ostensive definitions are but one type. But it 
is at best vacuous to claim that all nonlogical terms are names. There are indefinitely 
many grammatically different kinds of expressions, which fulfill different roles in a lan- 
guage and have different uses, given by the explanations of their meanings, which are 
in effect rules for their use. In the sense in which the Tractatus claimed that there is a 
connection - a meaning-endowing connection - between language and reality, there 
is no such connection. It was mistaken to suppose that a propositional sign is a fact, 
that only facts can represent facts, or that only "simple names" can represent simple 
objects. Far from the logical syntax of language having to mirror the logical forms of 
things, the different grammars of different languages are autonomous. They owe no 
homage to reality. They do not reflect language-independent metaphysical possibilities, 



determined by the essential nature of objects represented, but rather themselves deter- 
mine logical possibilities, i.e. what it makes sense to say. Empirical propositions are 
indeed characteristically (although not uniformly) bipolar, but the concept of a propo- 
sition is a family resemblance concept: there are many different kinds of proposition, 
which are not characterized by an essential nature, but by overlapping similarities. The 
concept of logical form which had informed the Tractatus is chimerical. For paraphrase 
into a canonical notation (as in Russell's theory of descriptions) is not an analysis of 
what is already present in the paraphrased proposition or thought but a redescription 
in a different form of representation. Logical form is no reflection of the logico- 
metaphysical forms of reality, since there is no such thing. 

Already in the Tractatus Wittgenstein had rejected the logicism in the philosophy of 
mathematics which Frege and Russell had endeavored unsuccessfully to prove. He 
denied that numbers were logical objects or reducible to classes. Mathematical 
propositions, he claimed, are not descriptions of possible states of affairs. Nor are they 
bipolar. They are, in effect, nonsensical pseudo-propositions; they do not have a sense 
consisting in their agreement and disagreement with the existence and nonexistence 
of states of affairs. Rather, they are substitution-rules for the transformation of 
one empirical proposition concerning magnitudes or quantities or spatial relations, 
etc. into another, and expressions of rules are not propositions. In the 1930s he wrote 
extensively about the foundations of mathematics. It is not possible here to do more 
than indicate briefly the general trajectory of his thought. He did not reject logicism in 
order to embrace what seemed to be the only alternatives, namely intuitionism and 
formalism. His fundamental claim is radical. With the liberalization in his concept of a 
proposition, he was now willing to speak of mathematical propositions. Nevertheless, 
they are radically unlike empirical propositions, and equally unlike logical ones. 
Mathematics is a system of interlocked propositions. As already implied in the Tractatus, 
the fundamental role of this system (but not of every proposition within it) is to 
constitute rules for the transformation of empirical propositions. An arithmetic equa- 
tion, such as 25 2 = 625, is a rule licensing the transformation of such an empirical 
proposition as "There are 25 boxes each containing 25 marbles" into the proposition 
"There are 625 marbles." A theorem of geometry is a norm of representation: a 
rule permitting the transformation of empirical propositions about shapes, distances, 
or spatial relations. Different geometries are not different theories about empirical space, 
which might turn out to be true or false. Nor are they different uninterpreted calculi. 
Rather, they are different grammars for the description of spatial relations. Proof by 
mathematics (e.g. in engineering) is wholly different from proof in mathematics. While 
a mathematical proposition is a rule, unless it is an axiom, it is not stipulated, but 
produced according to rules by a proof. Here we must distinguish proofs within a proof 
system, e.g. a computation, which is just "homework," as Wittgenstein put it, from 
proofs which extend mathematics by extending a proof system. Proofs that extend 
mathematics create new internal relations, modifying existing concepts by linking 
them with concepts with which they were hitherto unconnected, or connecting 
them with concepts in new ways - thus licensing novel transformations of appro- 
priate empirical (or other mathematical) propositions. Mathematics is concept 
formation. The propositions of mathematics determine the concepts they invoke. What 
we conceive of as mathematical necessity is at best a distorted reflection of the inter- 



nal relations within a proof system. Mathematics is a human creation, invented rather 
than discovered. 

The Philosophical Investigations 

Dismantling the Tractatus preoccupied Wittgenstein in the early 1930s. Gradually a 
new method and a wholly different conception of language, of linguistic meaning, and 
of the relation between language and reality emerged. It became clear that his neglect 
of questions in the philosophy of psychology in the Tractatus, which he had taken to be 
licensed by the anti-psychologism he took over from Frege, was unwarranted. For the 
concepts of linguistic meaning are bound up with the concepts of understanding, 
thinking, intending, and meaning something, and these pivotal notions demand philo- 
sophical elucidation. The new method also led to a new conception of philosophy itself 
- related to, but still importantly different from, the conception of philosophy advocated 
in the Tractatus. That in turn led to a different criticism of metaphysics. 

Successive efforts to compose a book laying forth his new ideas culminated in the 
composition of the Philosophical Investigations, Part 1, which was virtually completed 
by 1945/6. It is his masterwork. Despite some continuities of theme and negative 
conceptions, it stands in stark contrast not only to the sibylline style of the Tractatus 
but above all to its spirit. Where the Tractatus strove for a sublime insight into the 
language-independent essences of things, the Investigations proceeded by a quiet weigh- 
ing of linguistic facts in order to disentangle knots in our understanding. The Tractatus 
was possessed by a vision of the crystalline purity of the logical forms of thought, lan- 
guage, and the world, the Investigations was imbued with a sharpened awareness of the 
motley of language, the deceptive forms of which lead us into confusion. The Tractatus 
advocated conceptual geology, hoping to disclose the ineffable essences of things by 
depth analysis of language, the Investigations practiced conceptual topography, aiming 
to dissolve philosophical problems by a patient description of familiar linguistic 
facts. The Tractatus was the culmination of a tradition in western philosophy. The 
Investigations is virtually without precedent in the history of thought. 

Wittgenstein's later work, as he himself said, is not merely a stage in the continuous 
development of philosophy, but constitutes a "kink" in the development of thought 
comparable to that which occurred when Galileo invented dynamics; it was, in a sense, 
a new subject, an heir to what used to be called "philosophy." A new method had been 
discovered, and for the first time it would now be possible for there to be "skillful" 
philosophers - who would apply the method. The transition from the Tractatus to his 
later philosophy, as he wrote when his new ideas were dawning in 1929, is the transi- 
tion from the method of truth to the method of meaning. It is a transition from 
Wesensschau - putative insights into the nature or essence of things - to the clarifica- 
tion of conceptual connections in the grammar of our languages, with the purpose of 
disentangling knots in our thought. The conception of philosophy advocated in the 
Investigations has no precedent, although it is, in a qualified sense, anticipated by the 
Tractatus program for future philosophy. The philosophy of language is equally without 
ancestors: it is neither a form of idealist telementational linguistic theory (on the model 
of classical empiricism or de Saussure) nor a form of behaviorist linguistic theory, it is 
neither a realist truth-conditional semantics nor a form of "anti-realist" semantics. The 



philosophy of mind repudiates both dualism as well as mentalism on the one hand and 
logical behaviorism as well as physicalism on the other. The critique of metaphysics 
rests neither on Humean or verificationist grounds, nor does it resemble the Kantian 
critique of transcendent metaphysics. It is no wonder that Wittgenstein's later 
philosophy has been so frequently misunderstood and misinterpreted, for it can no 
more be located on received maps of philosophical possibilities than the North Star 
can be located on a terrestrial globe. 

The Investigations opens with a quotation from St. Augustine's autobiography in 
which he recounts the manner in which he assumes that he had learnt to speak. These 
unselfconscious, nonphilosophical reflections seemed to Wittgenstein to crystallize an 
important proto-picture of language, a pre-philosophical conception of its role and 
function, which informs a multitude of philosophical theories. According to this picture 
the essential role of words is to name things, and the essential role of sentences is to 
describe how things are. Words are connected to things by means of ostension. This 
proto-picture, which is akin to an unnoticed field of force unconsciously moulding the 
shape of sophisticated philosophical theories, is one root of extensive misconceptions 
in philosophy of language, logic, mathematics, and psychology. It is a muted leitmotif 
running through the book, and combating the influence of this picture is one of the 
central tasks of the book. For we are prone to think that corresponding to every name, 
or corresponding to every name on analysis, there must exist some thing: that nouns 
name objects, adjectives name properties, verbs name actions, that psychological 
expressions such as "pain" name psychological objects, and "believe," "want," 
"intend," "think," etc. name psychological states or processes, number words name 
numbers, and logical connectives name binary relations. We are inclined to believe that 
every declarative sentence describes something: that logical propositions describe rela- 
tions between thoughts, that mathematical propositions describe relations between 
numbers, that what we conceive of as metaphysical propositions describe necessary 
relations between ultimate categories of being, that psychological propositions in the 
first-person describe states of mind, and so on. But this is illusion. 

Philosophy of language 

The philosophy of language of the Investigations has a destructive and a constructive 
aspect. Its destructive aspect is concerned with undermining the conception of 
analysis that had informed the Tractatus and, more remotely, has characterized phil- 
osophy at least since the Cartesian and empiricist programs of analysis into simple 
natures and into simple ideas respectively. It aims to destroy the conception of a 
language as a calculus of meaning rules and the idea that the meaning or sense of a 
sentence is composed of the meanings of its constituent words and derivable from them, 
given their mode of combination. Hence too, it combats the ideal of determinacy of 
sense, and the thought that all expressions are either definable by analytic definition or 
are indefinables and hence explained by an ostensive definition, conceived of as linking 
language with reality and laying the foundations of language in simple objects given 
in experience. 

It has already been noted that the concepts of simple and complex are relative. Hence 
whether an A is complex or simple has to be determined by reference to criteria of 



simplicity and complexity laid down for As - if there are such criteria. But we com- 
monly confuse the absence of any criteria of complexity (since none have been laid 
down) with the satisfaction of criteria of simplicity. We are prone to think that an 
expression is complex if it is defined by analytic definition, and simple if it is explained 
by ostension. But analytic and ostensive definitions are neither exclusive nor exhaus- 
tive. We can explain what "circle" means by saying that this is a circle, or by saying that 
a circle is a locus of points equidistant from a given point. And we can explain what 
words mean by contextual paraphrase, contrastive paraphrase, exemplification, by a 
series of examples together with a similarity rider, by gesture, and so on. The meaning 
of a word is not an object for which a word stands or of which it is the name. Rather, 
it is what is given by an explanation of meaning, and an explanation of meaning is a 
rule for the use of the explanandum - a standard of correctness for its application. To 
ask for the meaning of a word is to ask how it is to be used. Indeed, the meaning of a 
word is (or, more cautiously, is determined by) its use. 

Ostensive definition is one legitimate manner of explaining the meanings of some 
words. It is not especially privileged: as argued, it does not "connect language with 
reality" or lay the foundations of language; it is only one rule for the use of the word 
in question, and it is as capable of being misunderstood as any other explanation of 
meaning. Many expressions do not have necessary and sufficient conditions of appli- 
cation. Among these are family-resemblance concepts, such as "game," which are 
explained by a series of examples and a similarity rider. (Even if someone comes up with 
a sharp definition of "game," that definition is not the rule by reference to which we 
have been applying the word "game" and by reference to which we would have justi- 
fied our use of the word.) Indeed, many of the pivotal concepts in philosophy, such as 
"language," "proposition," "number," "rule," "proof," as well as many psychological 
concepts, are family-resemblance concepts. Their extension is not determined by 
common properties, but by overlapping similarities - like the fibers in a rope. 

Since numerous kinds of expression are not explained in terms of necessary and suf- 
ficient conditions of application, the idea that vagueness is only a surface grammatical 
feature of language or that it must be an imperfection in language is awry. The Fregean 
demand for determinacy of sense was incoherent. For determinacy of sense is not 
merely the absence of vagueness, but the exclusion of the very possibility of vagueness: 
the exclusion, by a complete explanation of meaning, of every possibility of doubt in 
every conceivable circumstance. But there is no such thing. There is no absolute con- 
ception of completeness. The concepts of complete and incomplete are both relative and 
correlative. A complete explanation of meaning is an explanation which may be 
invoked as a standard of application in all normal contexts. Relative to that standard, 
explanations may be judged to be complete or incomplete. But we have no single ideal 
of exactness; what counts as exact or vague varies from context to context. Moreover, 
vagueness is not always a defect ("I ask him for a bread knife," Wittgenstein mocked, 
"and he gives me a razor blade because it is sharper"), and its occurrence is not logi- 
cally "contagious." 

The idea that the sense of a sentence is a function of the meanings of its constituents 
and their mode of composition is a distorted statement of the platitude that if one 
does not know what the words of a sentence mean or does not understand the way in 
which they are combined, then one will not understand what is said. The supposition 


that what a sentence means follows from an explanation of what its words mean, 
together with a specification of its structure, errs with regard to both meaning 
and understanding. The meaning of a sentence is no more composed of the meanings 
of its parts than a fact is composed of objects. The distinctions between sense and non- 
sense are not drawn once and for all by reference to circumstance-invariant features 
of type-sentences, but by reference to circumstance-dependent features of the use 
of token-sentences. Sentences of precisely the same form may have very different 
uses. Indeed, the forms of sentences, no matter whether in natural language or 
translated into a canonical notation of the predicate calculus, conceal rather than 
reveal their use. Moreover, understanding a sentence is not a process of deriving its 
meaning from anything. 

Little remains of analysis as previously understood. Philosophical problems are 
misunderstandings caused, among other things, by misleading analogies between 
forms of expressions with different uses. Some of these can be dissolved by paraphrase, 
as exemplified by Russell's Theory of Descriptions (see RUSSELL). But it was an illusion 
that there is anything like a final analysis of the forms of our language, let alone that 
analysis reveals the logical structure of the world. Instead of analysis as classically 
conceived, what is needed is a description of the uses of words that will illuminate 
philosophical confusion, and a rearrangement of familiar rules for the use of words 
which will make the grammar of the relevant expressions surveyable. For the main 
source of philosophical puzzlement and of misguided philosophical theories is our 
failure to command a clear view of the use of words and our consequent entanglement 
in the network of grammar. Connective analysis (the term is Strawsonian rather than 
Wittgensteinian), that is, a description of the conceptual connections and exclusions 
in the web of words, and therapeutic analysis (see below) replace reductive analysis. 
A sentence is completely analyzed, in the new sense, when its grammar is laid out 
completely clearly. 

A language is misrepresented if it is conceived to be a calculus of rules. More 
illuminating is the idea that it is a motley of language games. Language is indeed 
rule-governed, in the loose manner in which games are. Using sentences is compara- 
ble to making moves in a game, and a language can be fruitfully viewed as a motley of 
language games. The use of language is interwoven with the lives and practices 
of speakers, and is partly constitutive of their form of life. Training and teaching 
underpin the mastery of a language, and these presuppose shared reactive and 
behavioral propensities within a linguistic community. Words are like tools, and the 
diversity of their use is as great as that of different tools: hence masked by conceiving 
of them as essentially names of things, and concealed by their grammatical form. The 
greatest error of philosophers of his day, Wittgenstein remarked, is to attend to 
the forms of expressions rather than to their uses. Even declarative sentences are used 
for endlessly diverse purposes, of which describing is only one, and non-declarative 
sentences are misrepresented if taken to be analyzable into a force-indicative com- 
ponent (e.g. an assertion sign or interrogative sign) and a descriptive, truth-value 
bearing "sentence-radical." Moreover, the concept of description is itself non-uniform, 
for describing a scene is altogether unlike describing a dream, describing the impres- 
sion something made is unlike describing the item that made the impression, 
and describing what one intends is altogether unlike describing the execution of one's 



intention. These are logically distinct kinds of descriptions, with distinct kinds of 
grounds and consequences. 

Understanding is akin to an ability, not a state from which performance flows. The 
criteria for linguistic understanding are of three general kinds: correct use, giving a 
correct explanation of meaning in context, and responding appropriately to the use of 
an expression. Viewing explanations of meaning as rules for the use of words, the use 
of words as rule-following, and understanding as the mastery of the technique of the 
use of words requires that these concepts be tightly interlocked. And so they are. There 
is an internal relation between a rule and what counts as compliance with it, which is 
manifest not only in the interpretations one might give of the rule, but above all in the 
practice of acting in accordance with it, and in the critical practices of teaching the 
meanings of expressions, of correcting misapplications and mistaken explanations of 
meaning. Meaning is determined by use, it is given by an explanation of meaning, and 
it is what is understood when the meaning of an expression is understood. Not every 
difference in use is a difference in meaning, but every difference in meaning is a differ- 
ence in use. Wittgenstein's later philosophy of language is guided by this series of con- 
ceptual connections, the ramifications of which he explored in detail. 

Philosophy of mind 

Against prevailing tradition, Wittgenstein challenged the inner/outer picture of the 
mind, the conception of the mental as a "world" accessible to its subject by introspec- 
tion, the conception of introspection as inner perception, the idea that the capacity to 
say how things are with us "inwardly" is a form of knowledge (let alone a paradigm of 
self-knowledge), the thought that human behavior is "bare bodily movement," the 
notion that voluntary action is bodily movement caused by acts of will, the supposition 
that explanation of human behavior in terms of reasons and motives is causal, and the 
pervasive influence of the Augustinian picture of language that disposes one to think 
that psychological expressions are uniformly or even typically names of inner objects, 
events, processes, or states. His philosophy of mind and of action can be seen as 
providing a rigorous philosophical underpinning for the hermeneutic insistence on 
the autonomy of humanistic understanding and its categorial differentiation from 
understanding in the natural sciences. 

Psychological expressions are not names of entities which are directly observable 
only by the subject, and avowals of the inner are not descriptions of something visible 
only in a private peepshow. It is all too easy to think of psychological expressions as 
names of inner entities, and hence of assigning them meaning by private ostensive defi- 
nition. Wittgenstein's "private language arguments" are aimed at this misconception. 
There can be no inner, private, analogue of public ostensive definition. Sensations 
cannot fulfill the role of samples. So a pain cannot serve as a defining sample for the 
application of the word "pain." Concentrating one's attention upon one's pain is not a 
kind of pointing. Remembering a sensation presupposes and so cannot explain the 
meaning of a sensation-name, and the memory of a sensation cannot serve as an 
object of comparison for the application of a sensation-word. There is no such thing as 
applying an expression in accordance with a rule which is in principle incommunica- 
ble to anyone else. But the idea of defining a sensation word by reference to a 


sensation, conceived of as private and intended to function as a defining sample in an 
ostensive definition would be such a pseudo-rule - for which there could be no crite- 
rion of correct application. Whatever seemed to one to be right would be right, and that 
means that there is no such thing here as right or correct. 

Indeed, the very notion of privacy which informs Cartesian and empiricist concep- 
tions of the mental is misconceived. The mental was taken to be private in two senses: 
privately owned and epistemically private. Pains, for example, were held to be privately 
owned, i.e. only 1 can have my pain, another person cannot have my pain but only a 
qualitatively identical one. And only f can really know that 1 have a pain, others can 
only surmise that 1 do. Both of these claims are misconceived. To have a pain is not to 
own anything, any more than to have a birthday or a train to catch. The distinction 
between numerical and qualitative identity, which applies to substances, no more 
applies to pains (or mental images, thoughts or feelings) than it does to colors, ff A is 
red and B is red, then A and B are the same color; so too, if NN has a throbbing headache 
in his right temple and MM has a throbbing headache in his right temple, then NN and 
MM have the same headache - neither numerically the same, nor qualitatively the 
same, but just the same. To think that what differentiates my pain from yours is that 1 
have mine and you have yours is to transform the owner of the pain into a distin- 
guishing property of the pain - which is as absurd as claiming that two chairs cannot 
have the same color, since the color of this chair belongs to this chair and the colour of 
that chair belongs to that chair. 

The conception of epistemic privacy is equally awry. Far from the "inner" being a 
field of certain empirical knowledge possessed by the subject, which is better known 
than, and provides the foundations for, other kinds of empirical knowledge, first- 
person, present-tense psychological utterances are not generally expressions of knowl- 
edge at all. "1 know f am in pain" is either an emphatic or concessive assertion that 1 
am in pain, or philosophers' nonsense. In such cases, ignorance, doubt, mistake, 
misidentification, misrecognition are ruled out by grammar: we have no use for such 
forms of words as "I may be in pain, or I may not - 1 am not sure, I must find out." But 
we mistake the grammatical exclusion of ignorance, doubt, etc., for the presence of 
knowledge, certainty, correct identification, and recognition. Whereas they too are 
excluded as senseless in such cases as pain, and the use of the epistemic operators in 
other cases has a distinctive meaning; "I don't know what I want" or "I do not know 
what I believe" are not expressions of ignorance but of indecision. I do not need to 
look into my mind to find out what I want or believe, but to make it up. If I do not know 
what I believe about X, I need to examine the evidence, not my state of mind. The 
utterances "I am in pain," "I'm going to V," "I want G" are standardly employed as 
expressions or avowals (rather than descriptions) of pain, intention, or desire, and the 
utterance is a criterion for others to ascribe to the speaker the relevant psychological 

A criterion for the inner is logically (conceptually), as opposed to inductively, good 
evidence (justification) for ascribing to another an appropriate psychological predicate. 
Pain and pain behavior, or desire and conative behavior, are not analogically, induc- 
tively, or hypothetically connected. Rather, crying out in circumstances of injury, 
assuaging an injured limb, avoiding the cause of injury, etc. are non-inductive grounds 
for pain-ascriptions. Grasping the concept of pain involves recognizing such criteria as 



grounds for ascription of pain to another. The criteria for ascription of psychological 
predicates are partly constitutive of the relevant concepts. Psychological utterances or 
avowals of the inner are (in certain cases) learnt extensions of primitive behavior that 
manifests the inner. For example, an avowal of pain is grafted onto, and is a partial 
replacement of, a groan of pain; and while an utterance of pain is as groundless as a 
shriek of pain, it too constitutes a criterion for third-person ascriptions, ft is misguided 
to suggest that we can never know whether another is in pain. On the contrary, we 
often know with complete certainty. When someone severely injured screams with 
pain, just try to doubt whether he really is in pain! Self-knowledge is a hard won 
achievement, not gained merely by having toothache, wanting or thinking this or that, 
and being able to say so. Indeed, others often know and understand us better than we 
do ourselves. 

The mind is not a substance. It is not identical with the brain. It is not a private space, 
in which mental objects are paraded, disclosed to introspective vision. There is, to be 
sure, such a thing as introspection, but it is not inner perception. Rather it is a form of 
reflection on one's past, one reasons and motives, affections and attitudes. The third- 
person pronoun refers neither to the mind nor to the body, but to the person, the living 
human being. The first-person pronoun functions quite differently; here reference 
failure, misidentification, misrecognition, and indeterminacy of reference are stan- 
dardly excluded. "I" is at best a degenerate, limiting case of a referring expression, as a 
tautology is a limiting case of a proposition with a sense. 

Psychological predicates are neither predicable of the body nor of its parts. It is 
senseless to ascribe to the brain predicates applicable only to the whole creature, e.g. 
thinking, believing, wanting, or intending. For the criteria for the third-person ascrip- 
tion of such predicates are distinctive forms of behavior of the creature in the stream of 
life, and there is no such thing as a brain manifesting thought or thoughtlessness, belief 
or incredulity, desire or aversion, intention or inadvertence in what it does. Hence too, 
it makes no sense to ascribe thought or thoughtlessness, understanding, misunder- 
standing or failure of understanding to machines. Thought is essentially bound up with 
the sentient, affective, and conative functions of a being that has a welfare, is capable 
of desiring and suffering, can set itself goals and pursue them, and can hope to succeed 
or fear to fail in its projects. 

Human behavior that constitutes criteria for the ascription of psychological predi- 
cates is not "bare bodily movement, " from which we infer analogically or hypothetically 
their inner state or which we interpret as action. On the contrary, we see the pain in the 
face of the sufferer, hear the joy in the voice of a joyful person, perceive the affection 
in the looks of lovers. Pain, pace behaviorists, is not pain-behavior, any more than joy 
is the same as joyous behavior or love the same as a loving look. But the "inner" is not 
hidden behind the "outer" ; it may sometimes be concealed or suppressed (or it may just 
not be manifested). But if it is manifested, then it infuses the "outer," which is not bare 
bodily movement, but the actions and affective reactions of living sentient beings in the 
stream of life. These are not typically describable save in the rich vocabulary of the 

Human action is not movement caused by acts of will. There are such things as acts 
of will and great efforts of will, but they are unusual, and are not causal antecedents 
of action. There is such a thing as will power, but that is a matter of tenacity rather 



than a psychic analogue of muscle power. Voluntary actions are not actions, let alone 
movements, preceded by an act of will. Wanting and willing are not names of mental 
acts or processes, and "He V'ed because he wanted to" does not give a causal explana- 
tion of his action; on the contrary, it typically precludes one. Voluntary movement is 
action for which it makes sense to ask for agential reasons, which a person can decide 
to perform, try to execute, or be ordered to do. It is marked by lack of agential surprise, 
and the agent can be held responsible for it. 

A reason for action or for belief is a premise in reasoning. Hence it is no more 
causally related to the action for which it is a reason than the reasons for a belief are 
causally related to the conclusion which they support. A person's reason is given by 
specifying the reasoning he went through antecedently to acting or the reasoning he 
could have gone through and is willing to give ex post actu. Reasons, unlike causes, 
justify or purport to justify that for which they are reasons. A person's avowal of a 
reason for his action, unlike his typical assertion of a cause of some event, is not 
a hypothesis. Unlike the assertion of a cause, in the standard case of an avowal of a 
reason, there is no room for mistake. What makes the connection between the reason 
and the action is the agent's avowal itself. In avowing a reason, the agent typically takes 
responsibility for his action viewed under the aspect of the avowed reason. 

The critique of metaphysics and nature of philosophy 

The Tractatus program for future philosophy advocated a non-cognitive conception of 
philosophy, denying that there could be any philosophical propositions, a fortiori any 
philosophical knowledge. Philosophy should be an activity of elucidation by analysis. 
Although philosophy was deprived of the possibility of stating essential truths about 
the natures of things, these very truths were held to be shown by the well-formed propo- 
sitions of a language, and arriving at a correct logical point of view would include 
apprehension and appreciation of what cannot be said but shows itself (including 
truths of ethics and aesthetics). 

The later conception of philosophy adhered to the radical non-cognitivism, but 
rejected the doctrine of linguistically manifest ineffabilia. There are indeed no philo- 
sophical truths. What appear as such, and what were construed by the Tractatus as an 
attempt to say what can only be shown, are in effect expressions of rules for the use of 
expressions in the misleading guise of metaphysical descriptions of the nature of 
things. So the portentous, apparently metaphysical, claim that the world consists 
of facts not of things amounts to the grammatical statement that a description of the 
world consists of a statement of facts and not a list of things. And that in turn is just 
a rule for the use of the expression "a description of the world." Insofar as metaphysics 
is conceived to be the quest for knowledge of the necessary forms and structures of the 
world or of the mind, it is chimerical. All that can be gleaned from these barren 
fields are grammatical propositions, that is, expressions of rules for the use of words. 
There are no such things as "necessary facts," and sentences such as "red is a color," 
or "space is three dimensional" are in effect rules. If something is said to be red, then 
it can be said to be colored; if something is in space, then its location is given by three 
coordinates; and so on. Similarly, apparently synthetic a priori truths, such as "Black 
is darker than white" or "Red is more like orange than it is like yellow," are not insights 


into language-independent necessities in the world, but expressions of rules that are 
partly constitutive of the meanings of the constituent expressions. For any ordered 
pair of samples which can be used to define "black" and "white" ostensively can also 
be used to define the relation "darker than." So if a is black and b is white, it follows 
without more ado that a is darker than b. If a is red, b orange, and c yellow, then 
a is more akin to b in color than to c - one need not look to see. The apparently 
metaphysical proposition is in fact an inference rule, which is partly constitutive of the 
meanings of the constituent terms. What appear to be descriptions of objective 
necessities in the world are merely the shadows cast by the rules for the use of color 
predicates and relations. 

Similarly, "cannot" and "must" in putatively metaphysical propositions mask rules 
for the use of words. "You cannot travel back in time" or "You cannot count through 
all the cardinal numbers" look like 'An iron nail cannot scratch glass," but they are 
not. Experience teaches that iron cannot scratch glass. But it is not experience that 
teaches that one cannot travel in time, rather, it is grammar that stipulates that the 
form of words "I travelled back to last year" has no use; nothing counts as travelling 
backwards in time. "Cannot" in metaphysics is not about human frailty, but is an 
expression of a convention. "You cannot count through all the cardinal numbers" is 
an expression of a grammatical rule which excludes the phrase "counting through all 
the cardinal numbers" from the language, ft does not say that there is something we 
cannot do, but rather that there is no such thing to do. Similarly, "must" in metaphysics 
signifies not an objective necessity in reality, but a commitment to a form of represen- 
tation. "Every event has a cause" is a true or false empirical generalization. "Every event 
must have a cause" is an expression of a commitment not to call anything "an event" 
unless it has a (known or unknown) cause. 

There are no theories in philosophy, for there can be nothing hypothetico-deductive 
about the determination of the bounds of sense, nor can it be merely probable that such 
and such a philosophical pronouncement makes no sense. And we do not need to wait 
upon future confirmation to determine with certainty that it makes no sense. Hence 
too, there is no philosophical knowledge comparable to knowledge in the sciences. If 
anyone were to advance theses in philosophy, everyone would agree with them: for 
example, "Can one step twice into the same river?" - "Yes." Indeed, there are no expla- 
nations in philosophy in the sense in which there are in the sciences, for the methods 
of philosophy are purely descriptive, and not methods of hypothesis formation. 

The purified non-cognitivism of the Investigations has two aspects. On the one hand, 
philosophy is a quest for a surveyable representation of a segment of our language with 
the purpose of solving or dissolving philosophical perplexity. On the other hand, 
philosophy is a cure for diseases of the understanding. Philosophical problems are 
conceptual, hence a priori and not empirical. They can be neither solved nor advanced 
by new information or scientific discoveries, although scientific discoveries may, and 
often do, raise fresh conceptual puzzles and generate new confusions. Conceptual prob- 
lems may concern novel concept-formation or existing conceptual structures and 
relations. The former are exemplified by mathematics, the latter by philosophy. The task 
of philosophy is to resolve conceptual questions arising out of our existing forms of 
representation, to clarify conceptual confusions that result from entanglement in the 
web of the grammar of our language. Philosophy is not a contribution to human 


knowledge but to human understanding - an understanding of our forms of repre- 
sentation and their articulations, an overview of the forms of our thought. 

The main source of philosophical puzzlement and of misconceived philosophical 
theories is our failure to command a clear view of the uses of words. The grammar of 
our language is lacking in surveyability, for expressions with very different uses have 
similar surface grammars: "I meant" looks akin to "I pointed," "I have a pain" to "I 
have a pin," "He is thinking" appears akin to "He is talking," "to have a mind" looks 
like "to have a brain," "2 is greater than 1" seems akin to "Jack is taller than Jill." Hence 
we misconstrue the meanings of expressions in our philosophical reflections. We think 
of meaning something or someone as a mental act or activity of attaching signs to 
objects, take pain to be a kind of object inalienably possessed by the sufferer, imagine 
that the mind is identical with the brain, assume that statements of numerical 
inequalities are descriptions, and so on. 

What is needed is a perspicuous representation of the segment of grammar that 
bears on the problem with which we are confronted. It enables us to see differences 
between concepts that are obscured by the misleadingly similar grammatical forms of 
expressions. For this no new discoveries are necessary or possible - only the description 
of grammar, the clarification and arrangement of familiar rules for the use of words. 
We must remind ourselves of what we already know perfectly well, namely how expres- 
sions, the use of which we have already mastered, are indeed used. To be sure, these 
rules must then be arranged in such a manner as to shed light upon the difficulty in 
question. The rules that concern the philosopher are different from those that concern 
the grammarian, and the ordering of rules by the philosopher is very different from the 
ordering sought by the grammarian, for their purposes are quite distinct. A perspicu- 
ous representation of a fragment of grammar will enable us to find our way around the 
relevant part of the grammatical network without stumbling into conceptual confu- 
sion. In philosophy, unlike in the sciences, all the information is already at hand - in 
our knowledge of our language. The problems of philosophy, unlike those of science, 
are completely solvable. Failure to solve them is due to philosophers' failure to arrange 
the grammatical facts in such a way that the problems disappear. 

Complementary to the conception of philosophy as the quest for a surveyable rep- 
resentation of segments of our language that give rise to conceptual perplexity and con- 
fusion is the conception of philosophy as therapeutic. The philosopher's treatment of a 
question is like the treatment of an illness. One should not try to terminate a disease of 
thought, either by dogmatism or by the substitution of a technical concept for the prob- 
lematic one that causes confusion (as Carnap did with his method of "explication"), for 
slow cure is all important. Every deep philosophical confusion has many different roots, 
and each must be dug up and examined. Every deeply misconceived answer to a philo- 
sophical problem that mesmerizes us and holds us in a vice has many facets, and each 
must be separately surveyed. Wittgenstein sometimes compared his new methods of 
philosophical clarification with psychoanalysis. Philosophical theories are latent non- 
sense; the task of the philosopher is to transform them into patent nonsense. Like the 
psychoanalyst, the philosopher aims to give the afflicted insight into their own under- 
standing and misunderstanding. 

Philosophy is categorially distinct from the sciences. Since there is no philosophical 
knowledge and there are no licit theories in philosophy, there can be no progress in the 



sense in which there is in the sciences. For there is no accumulation of knowledge, 
no generation of ever richer explanatory theories, no refinement of instrumentation 
making possible ever more accurate measurement and observation. But there can be 
progress in another sense, namely in clarification of conceptual structures, in drawing, 
refining and sharpening distinctions, in destroying conceptual illusions and in eradi- 
cating conceptual confusions. However, since there is no way of predicting future forms 
of entanglement in the web of language, the task of philosophy never ends. 

Wittgenstein's place in postwar analytic philosophy 

Is Wittgenstein's later philosophy a form of analytic philosophy? The concept of ana- 
lytic philosophy is neither sharply defined nor uncontested. If one takes the concept of 
analysis narrowly, connecting it primarily with decompositional analysis, with reduc- 
tion and logical construction, then one will be inclined to associate analytic philosophy 
primarily with a variety of forms of philosophy that flourished in the first half of the 
twentieth century. One will also be prone to associate the movement with a profound 
interest in, and ingenious philosophical use of, the calculi of formal logic, and, in some 
cases, in the devising of formal or semi-formal languages to replace the apparently 
defective natural languages for philosophical purposes. Moore and Russell, the young 
Wittgenstein of the Tractatus, the Cambridge analysts of the early interwar years, and 
the logical positivists will then be one's paradigmatic analytic philosophers. 

Thus construed, it is clear that it would be at best misleading to characterize the later 
Wittgenstein as an analytic philosopher at all. But it would be perverse to construe ana- 
lytic philosophy thus. The term "analytic philosophy" was a latecomer upon the scene, 
and the Oxford philosophers of the postwar era had no qualms in characterizing their 
work as analytic philosophy and their methods as conceptual analysis. This did not 
imply that they were dedicated to reductive analysis and logical construction. Indeed, 
they repudiated them. What it implied was a looser sense of "analysis": the description 
of the conceptual connections and articulations of salient elements in our conceptual 
scheme. In this sense, to be sure, the later Wittgenstein was an analytic philosopher, 
and said as much. For, he claimed, a proposition is "fully analyzed" when its grammar 
has been completely laid bare. Taken in this broader sense, analytic philosophy con- 
tinued after 1945 in a new and distinctive form. It was dominated by Oxford rather 
than Cambridge philosophers, although Wittgenstein's philosophy, transmitted to 
Oxford largely by word of mouth before 1953, was a primary influence upon them 


This postwar phase of analytic philosophy lasted for a quarter of a century. It was 
not a "school" and, unlike the Vienna Circle, issued no manifestos. It was united by its 
conception of philosophy as an a priori conceptual investigation, contributing to 
human understanding rather than to human knowledge, hence wholly unlike the sci- 
ences. There was consensus that the methodical examination of the use of the relevant 
words is a sine qua non of any serious philosophical investigation. Analytic philosophy 
of language flourished, as did analytic epistemology; so too did analytic philosophy of 
psychology and philosophy of action. Paths pioneered by Wittgenstein were followed 
and refined. But other branches of analytic philosophy, such as analytic jurisprudence, 
analytic aesthetics, analytic philosophy of history and the social sciences, which had 



been of little or no concern to him, were also developed, often in a manner which bore 
the marks of his influence. 

This phase of analytic philosophy waned in the 1970s, and Wittgenstein's influence 
declined. Whether the forms of philosophy that succeeded it are to be counted as yet 
another phase of analytic philosophy or as symptoms of its final demise is something 
that will become clearer only with the passing of time. What is, however, clear, is that 
Wittgenstein dominated the forms of analytic philosophy from the 1920s until the 
1970s, ineradicably impressing his thought upon twentieth-century philosophy. 


1 For example, Alice Ambrose, Elizabeth Anscombe, Max Black, Richard Braithwaite, Karl 
Britton, Peter Geach, Austin Duncan-Jones, Casimir Lewy, Margaret MacDonald, Norman 
Malcolm, G. A. Paul, Rush Rhees, Stephen Toulmin, John Wisdom, Georg Henrik von Wright. 

2 The other influences upon his thought which he cited retrospectively in 1931 were Karl 
Kraus, Adolf Loos, Paul Ernst, and Otto Weininger. In later years he made much use of 
James's The Principles of Psychology, which he viewed as a useful source of interesting philo- 
sophical confusions - hence not so much an influence upon his own ideas as a stimulus to 

3 To this must be added a large quantity of dictations he gave to Friedrich Waismann for the 
projected joint work Logik, Sprache, Philosophic which was intended as the first volume of the 
Vienna Circle's series Schriften zur Wissenschaftliches Weltauffassung, that volume itself, pub- 
lished in English under the title Principles of Linguistic Philosophy, and Waismann's notes of 
conversations with Wittgenstein published under the title Wittgenstein and the Vienna Circle. 

4 This contrasts with the Fregean and Russellian conception of the nature of the proposition. 
Frege held that propositions of natural language may lack a truth-value, although they 
express a sense. In his logically ideal language, Begriffsschrift, every proposition must be 
bivalent (but not bipolar), i.e. either true or false. Russell held propositions to be bivalent. 


Works by Wittgenstein 

1953: Philosophical Investigations, ed. G.E. M. Anscombe andR. Rhees, trans. G. E. M. Anscombe, 

Oxford: Blackwell Publishers. 
1958: The Blue and Brown Books, Oxford: Blackwell Publishers. 
1961a: Notebooks 1914-16, ed. G. H. von Wright and G. E. M. Anscombe, trans. G. E. M. 

Anscombe, Oxford: Blackwell Publishers. 
1961b: Tractatus Logico-Philosophicus, trans, by D. F. Pears and B. F. McGuinness, London: 

Routledge and Kegan Paul. 
1967: Zettel, ed. G. E. M. Anscombe and G. H. von Wright, trans. G. E. M. Anscombe, Oxford: 

Blackwell Publishers. 
1969: On Certainty, ed. G. E. M. Anscombe and G. H. von Wright, trans. D. Paul and G. E. M. 

Anscombe, Oxford: Blackwell Publishers. 
1970: Lectures and Conversations on Aesthetics, Psychology, and Religious Beliefs, compiled from 

notes by Y. Smythies, R. Rhees, and J. Taylor, ed. C. Barrett, Oxford: Blackwell Publishers. 
1971: ProtoTractatus: An Early Version of Tractatus Logico-Philosophicus, ed. B. F. McGuinness, 

T. Nyberg, and G. H. von Wright, trans. D. F. Pears and B. F. McGuinness, London: Routledge 

and Kegan Paul. 



1974: Philosophical Grammar, ed. R. Rhees, trans. A. J. P. Kenny, Oxford: Blackwell Publishers. 
1975: Philosophical Remarks, ed. R. Rhees, trans. R. Hargreaves and R. White, Oxford: Blackwell 

1976: Wittgenstein's Lectures on the Foundations of Mathematics, Cambridge 1 939, ed. C. Diamond, 

Sussex: Harvester Press. 
1978: Remarks on the Foundations of Mathematics, ed. G. H. von Wright, R. Rhees, and G. E. M. 

Anscombe, trans. G. E. M. Anscombe, rev. ed., Oxford: Blackwell Publishers. 
1979a: Ludwig Wittgenstein undder Wiener Kreis, shorthand notes recorded by F. Waismann, ed. 

B. F. McGuinness, Oxford: Blackwell Publishers, 1967. (English translation, Wittgenstein and 
the Vienna Circle, Oxford: Blackwell Publishers.) 

19 79b: Wittgenstein's Lectures, Cambridge 1 932-35, from the Notes of Alice Ambrose and Margaret 

MacDonald, ed. Alice Ambrose, Oxford: Blackwell Publishers. 
1980a: Culture and Value, ed. G. H. von Wright in collaboration with H. Nyman, trans. P. Winch, 

Oxford: Blackwell Publishers. 
1980b: Remarks on the Philosophy of Psychology, vol. I, ed. G. E. M. Anscombe and G. H. von 

Wright, trans. G. E. M. Anscombe, Oxford: Blackwell Publishers. 
19 80c: Remarks on the Philosophy of Psychology, vol. II, ed. G. H. von Wright and H. Nyman, trans. 

C. G. Luckhardt and M. A. E. Aue, Oxford: Blackwell Publishers. 

1980d: Wittgenstein's Lectures, Cambridge 1930-32, from the Notes of John King and Desmond Lee, 

ed. Desmond Lee, Oxford: Blackwell Publishers. 
1982: Last Writings on the Philosophy of Psychology, vol. I, ed. G. H. von Wright and H. Nyman, 

trans. C. G. Luckhardt and M. A. E. Aue, Oxford: Blackwell Publishers. 
1988: Wittgenstein's Lectures on Philosophical Psychology 1946-47, notes by P. T. Geach, K. J. 

Shah, and A. C. Jackson, ed. P. T. Geach, Hemel Hempstead: Harvester Wheatsheaf. 
1992: Last Writings on the Philosophy of Psychology, vol. II, ed. G. H. von Wright and H. Nyman, 

trans. C. G. Luckhardt and M. A. E. Aue, Oxford: Blackwell Publishers. 
1993: Ludwig Wittgenstein: Philosophical Occasions 1912-1951, ed. James Klagge and Alfred 

Normann, Indianapolis: Hackett. 

Works by other authors 

Baker, G. P. and Hacker, P. M. S. (1980-96), An Analytic Commentary on the Philosophical 
Investigations, 4 vols, Oxford: Blackwell Publishers; vol. 1, Wittgenstein: Meaning and 
Understanding (Baker and Hacker); vol. 2, Wittgenstein: Rules, Grammar and Necessity (Baker 
and Hacker); vol. 3, Wittgenstein: Meaning and Mind (Hacker); and vol. 4, Wittgenstein: Mind 
and Will (Hacker). 

Clock, H. J. (1996) A Wittgenstein Dictionary, Oxford: Blackwell Publishers. 

Hacker, P. M. S. (1986) Insight and Illusion: Themes in the Philosophy of Wittgenstein, rev. edn., 
Oxford: Oxford University Press and Bristol: Thoemmes Press. 

Kenny, A. J. P. (19 73) Wittgenstein, London: Penguin Books. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

Rudolf Carnap (1891-1970) 


Rudolf Carnap, pre-eminent member of the Vienna Circle, was one of the most influ- 
ential figures of twentieth-century analytic philosophy. The Vienna Circle was respon- 
sible for promulgating a set of doctrines (initially in the 1920s) which came to be 
known as logical positivism or logical empiricism. This set of doctrines provides the 
point of departure for most subsequent developments in the philosophy of science. 
Consequently Carnap must be regarded as one of the most important philosophers of 
science of this century. Nevertheless, his most lasting positive contributions were in the 
philosophy of logic and mathematics and the philosophy of language. Meanwhile, his 
systematic but ultimately unsuccessful attempt to construct an inductive logic has been 
equally influential since its failure has convinced most philosophers that such a project 
must fail. 

Carnap was born in 1891 in Ronsdorf, near Bremen, and now incorporated into the 
city of Wuppertal, in Germany. 1 In early childhood he was educated at home by his 
mother, Anna Carnap (nee Dorp fe Id), who had been a schoolteacher. From 1898, he 
attended the Gymnasium at Barmen, where the family moved after his father's death 
that year. In school, Carnap's chief interests were in mathematics and Latin. From 
1910 to 1914 Carnap studied at the universities of Jena and Freiburg, concentrating 
first on philosophy and mathematics and, later, on philosophy and physics. Among his 
teachers in Jena were Bruno Bauch, a prominent neo-Kantian, and Gottlob Frege, a 
founder of the modern theory of quantification in logic. Bauch impressed upon him the 
power of Kant's conception that the geometrical structure of space was determined by 
the form of pure intuition. Though Carnap was impressed by Frege's ongoing philo- 
sophical projects, his real (and lasting) influence only came later through a study of his 
writings (see FREGE). Carnap's formal intellectual work was interrupted between 1914 
and 1918 while he did military service during World War I. His political views had 
already been of a mildly socialist/pacifist nature. The horrors of the war served to make 
them more explicit and more conscious, and to codify them somewhat more rigorously. 


After the war, Carnap returned to Jena to begin research. His contacts with Hans 
Reichenbach and others pursuing philosophy informed by current science began 



during this period. In 1919 he read Whitehead and Russell's Principia Mathematica and 
was deeply influenced by the clarity of thought that could apparently be achieved 
through symbolization. He began the construction of a putative axiom system for a 
physical theory of space-time. The physicists - represented by Max Wien, head of the 
Institute of Physics at the University of Jena - were convinced that the project did not 
belong in physics. Meanwhile, Bauch was equally certain that it did not belong in phi- 
losophy. This incident was instrumental in convincing Carnap of the institutional dif- 
ficulties faced in Germany of doing interdisciplinary work that bridged the chasm 
between philosophy and the natural sciences. It also probably helped generate the atti- 
tude that later led the logical empiricists to dismiss much of traditional philosophy espe- 
cially metaphysics. By this point in his intellectual development (the early 1920s) 
Carnap was already a committed empiricist who, nevertheless, accepted both the ana- 
lyticity of logic and mathematics and the Frege-Russell thesis of logicism which 
required that mathematics be formally constructed and derived from logic. 

Faced with this lack of enthusiasm for his original project in Jena, Carnap aban- 
doned it to write a dissertation on the philosophical foundations of geometry, which 
was subsequently published as Der Raum (1922). A fundamentally neo-Kantian work, 
it included a discussion of "intuitive space," determined by pure intuition, independent 
of all contingent experience, and distinct from both mathematical (or abstract) space 
and physical space. However, in contrast to Kant, Carnap restricted what could be 
grasped by pure intuition to some topological properties of space; metric properties and 
even the dimensionality of space were regarded as empirical matters. In agreement 
with Helmholtz and Moritz Schlick (a physicist-turned-philosopher, and founder of the 
Vienna Circle - see below), the geometry of physical space was also regarded as an 
empirical matter. Carnap included a discussion of the role of non-Euclidean geometry 
in Einstein's General Relativity Theory. By distinguishing between intuitive, mathe- 
matical, and physical spaces, Carnap attempted to resolve the apparent differences 
between philosophers, mathematicians, and physicists by assigning the disputing 
camps to different discursive domains. In retrospect, this move heralded what later 
became the most salient features of Carnap's philosophical work: tolerance for diverse 
points of view (so long as they met stringent criteria of clarity and rigor) and an assign- 
ment of these viewpoints to different realms, the choice between which is to be resolved 
not by philosophically substantive (for instance, epistemological) criteria but by prag- 
matic ones. 

The constructionist phase 

During the winter of 1921, Carnap read Russell's Our Knowledge of the External World 
(1914). According to Carnap's intellectual autobiography (1963a), this work led him, 
between 1922 and 1925, to begin the analysis that culminated in Der logische Aufbau 
der Welt (1967), which is usually regarded as Carnap's first major work. The purpose 
of the Aufbau was to construct the everyday world from a phenomenalist basis. This is 
an epistemological choice (§§54, 58). 2 Carnap distinguished between four domains 
of objects: autopsychological, physical, heteropsychological, and cultural (§58). The 
first of these consists of objects of an individual's own psychology; the second of physi- 
cal entities (Carnap does not distinguish between everyday material objects and the 

9 5 


abstract entities of theoretical physics); the third consists of the objects of some other 
individual's psychology; and the fourth of cultural objects {geistige Gegenstdnde), which 
include historical and sociological phenomena. 

From Carnap's point of view, "[a]n object ... is called epistemically primary relative 
to another one ... if the second one is recognized through the mediation of the first 
and thus presupposes, for its recognition, the recognition of the first" (§54). 
Autopsychological objects are epistemically primary relative to the others in this sense. 
Moreover, physical objects are epistemically primary to heteropsychological ones 
because the latter can only be recognized through the mediation of the former: an 
expression on a face, a reading in an instrument, etc. Finally, heteropsychological 
objects are epistemically primary relative to cultural ones for the same reason. 

The main task of the Aujbau is construction, which Carnap conceives of as the con- 
verse of what he regarded as reduction (which is far from what was then - or is now - 
conceived of as "reduction" in Anglophone philosophy): 

an object is 'reducible' to others . . . if all statements about it can be translated into statements 
which speak only about these other objects. ... By constructing a concept from other concepts, 
we shall mean the indication of its "constructional definition" on the basis of other con- 
cepts. By a constructional definition of the concept a on the basis of the concepts b and c, we 
mean a rule of translation which gives a general indication how any propositional func- 
tion in which a occurs may be transformed into a coextensive propositional function in 
which a no longer occurs, but only b and c. If a concept is reducible to others, then it must 
indeed be possible to construct it from them. (§3 5) 

However, construction and reduction present different formal problems because, except 
in some degenerate cases (such as explicit definition), the transformations in the two 
directions may not have any simple explicit relation to each other. The question of 
reducibility/constructibility is distinct from that of epistemic primacy, fn an important 
innovation in an empiricist context, Carnap argues that both the autopsychological and 
physical domains can be reduced to each other (in his sense). Thus, at the formal level, 
either could serve as the basis of the construction. It is epistemic primacy that dictates 
the choice of the former. 

Carnap's task, ultimately, is to set up a constructional system that will allow the 
construction of the cultural domain from the autopsychological through the two inter- 
mediate domains. In the Aufbau, there are only informal discussions of how the 
last two stages of such a construction are to be executed. Only the construction of 
the physical from the autopsychological is fully treated formally. As the basic units of 
the constructional system Carnap chose what he calls "elementary experiences" 
(Elementarerlebnisse) {elex). 3. These are supposed to be instantaneous cross-sections of 
the stream of experience - or at least bits of that stream in the smallest perceivable unit 
of time - that are incapable of further analysis. The only primitive relation that Carnap 
introduces is "recollection of similarity" (Rs). (In the formal development of the system, 
Rs is introduced first and the elex are defined as the field of Rs.) The asymmetry of Rs 
is eventually exploited by Carnap to introduce temporal ordering. 

Since the elex are elementary, they cannot be further analyzed to define what would 
be regarded as constituent qualities of them, such as partial sensations or intensity 



components of a sensation. Had the elex not been elementary, Carnap could have used 
"proper analysis" to define such qualities by isolating the individuals into classes on the 
basis of having a certain (symmetric) relationship with each other. Carnap defines the 
process of "quasi-analysis" to be formally analogous to proper analysis but only defin- 
ing "quasi-characteristics" or "quasi-constituents" because the elex are unanalyzable. 4 
Quasi-analysis based on the relation "part similarity" (Ps), itself defined from Rs, is the 
central technique of the Aufbau. It is used eventually to define sense classes and, then, 
the visual sense, visual field places, the spatial order of the visual field, the order of 
colors and, eventually, sensations. Thus the physical domain is constructed out of the 
autopsychological. Carnap's accounts of the construction between the other two 
domains remain promissory sketches. 

Carnap was aware that there were unresolved technical problems with his con- 
struction of the physical from the autopsychological, though he probably underesti- 
mated the seriousness of these problems. The systematic problems are that when a 
quality is defined as a class selected by quasi-analysis on the basis of a relation: (1) two 
(different) qualities that happen always to occur together (say, red and hot) will never 
be separated; and (2) quality classes may emerge in which any two members bear some 
required relation to each other but there may yet be no relation that holds between all 
members of the class. Carnap's response to these problems was extra-systematic: in the 
complicated construction of our world from our elex, he hoped that such examples 
would never or only very rarely arise.' Nevertheless, because of these problems, and 
because the other constructions are not carried out, the attitude of the Aufbau is 
tentative and exploratory: the constructional system is presented as essentially 
unfinished. 6 

By this point of his intellectual development, Carnap had not only fully endorsed the 
logicism of the Principia, but also the form that Whitehead and Russell had given to 
logic (that is, the ramified theory of types including the axioms of infinity and reducibil- 
ity) in that work. However, Poincare also emerges as a major influence during this 
period. Carnap did considerable work on the conceptual foundations of physics in the 
1920s and some of this work - in particular, his analysis of the relationship between 
causal determination and the structure of space - shows strong conventionalist atti- 
tudes (Carnap 1924; see also Carnap 1923 and 1926). 

Viennese positivism 

In 1926, at Schlick's invitation, Carnap moved to Vienna to become a Privatdozent 
(instructor) in philosophy at the University of Vienna for the next five years. An early 
version of the Aufbau served as his Habilitationsschrift. He was welcomed into the Vienna 
Circle, a scientific philosophy discussion group organized by (and centered around) 
Schlick, who had occupied the Chair for Philosophy of the Inductive Sciences since 
1922. In the meetings of the Vienna Circle the typescript of the Aufbau was read and 
discussed. What Carnap seems to have found most congenial in the Circle - besides its 
members' concern for science and competence in modern logic - was their rejection of 
traditional metaphysics. Over the years, besides Carnap and Schlick, the Circle included 
Herbert Feigl, Kurt Godel, Hans Hahn, Karl Menger, Otto Neurath, and Friedrich 
Waismann, though Godel would later claim that he had little sympathy for the 



anti-metaphysical position of the other members. The meetings of the Circle were char- 
acterized by open, intensely critical, discussion with no tolerance for ambiguity of for- 
mulation or lack of rigor in demonstration. The members of the Circle believed that 
philosophy was a collective enterprise in which progress could be made. 

These attitudes, even more than any canonical set of positions, characterized the 
philosophical movement, initially known as logical positivism and, later, as logical 
empiricism, that emerged from the work of the members of the Circle and a few others, 
especially Reichenbach. However, besides rejecting traditional metaphysics, most 
members of the Circle accepted logicism and a sharp distinction between analytic and 
synthetic truths. The analytic was identified with the a priori; the synthetic with the a 
posteriori. A. J. Ayer, who attended some meetings of the Circle in 1933 (after Carnap 
had left - see below) returned to London and published Language, Truth and Logic in 
1936 (see AYER). This short book did much to popularize the views of the Vienna Circle 
among Anglophone philosophers though it lacks the sophistication that is found in the 
writings of the members of the Circle, particularly Carnap. 

Under Neurath's influence, during his Vienna years, Carnap abandoned the phe- 
nomenalist language he had preferred in the Aufbau and came to accept physicalism. 
The epistemically privileged language is one in which sentences reporting empirical 
knowledge of the world ("protocol sentences") employ terms referring to material 
bodies and their observable properties. From Carnap's point of view, the chief advan- 
tage of a physicalist language is its intersubjectivity. Physicalism, moreover, came hand- 
in-hand with the thesis of the "unity of science," that is, that the different empirical 
sciences (including the social sciences) were merely different branches of a single 
unified science. To defend this thesis, it had to be demonstrated that psychology could 
be based on a physicalist language. In an important paper only published somewhat 
later, Carnap (1934b) attempted that demonstration. Carnap's adoption of physicalism 
was final; he never went back to a phenomenalist language. However, what he meant 
by "physicalism" underwent radical transformations over the years. By the end of his 
life, it meant no more than the adoption of a non-solipsistic language, that is, one in 
which intersubjective communication is possible (Carnap 1963b). 

In the Vienna Circle, Wittgenstein's Tractatus was discussed in detail. Carnap found 
Wittgenstein's rejection of metaphysics concordant with the views he had developed 
independently. Partly because of Wittgenstein's influence on some members of the 
Circle (though not Carnap), the rejection of metaphysics took the form of an assertion 
that the sentences of metaphysics are meaningless in the sense of being devoid of cog- 
nitive content. Moreover, the decision whether a sentence is meaningful was to be made 
on the basis of the principle of verifiability, which claims that the meaning of a sen- 
tence is given by the conditions of its (potential) verification. Observation terms are 
directly meaningful on this account. Theoretical terms only acquire meaning through 
explicit definition from observation terms. Carnap's major innovation in these discus- 
sions within the Circle was to suggest that even the thesis of realism - asserting the 
"reality" of the external world - is also meaningless, a position not shared by Schlick, 
Neurath, or Reichenbach. Problems generated by meaningless questions became the 
celebrated "pseudo-problems" of philosophy (Carnap 1967). 

Wittgenstein's principle of verifiability posed fairly obvious problems in any scien- 
tific context. No universal generalization can ever be verified. Perhaps independently, 



Karl Popper perceived the same problem (see POPPER). This led him to replace the 
requirement of verifiability with that of falsifiability, though only as a criterion to 
demarcate science from metaphysics, and not as one also to be used to demarcate mean- 
ingful from meaningless claims. It is also unclear what the status of the principle itself 
is, that is, whether it is meaningful by its own criterion of meaningfulness. Carnap, as 
well as other members of the Vienna Circle including Hahn and Neurath, realized that 
a weaker criterion of meaningfulness was necessary. Thus began the program of the 
"liberalization of empiricism." There was no unanimity within the Vienna Circle on this 
point. The differences between the members are sometimes described as those between 
a conservative "right" wing, led by Schlick and Waismann, which rejected the liberali- 
zation of empiricism, and the epistemological anti-foundationalism that is involved in 
the move to physicalism; and a radical "left" wing, led by Neurath and Carnap, which 
endorsed the opposite views. The "left" wing also emphasized fallibilism and pragmat- 
ics; Carnap went far enough along this line to suggest that empiricism itself was a 
proposal to be accepted on pragmatic grounds. This difference also reflected political 
attitudes insofar as Neurath, and to a lesser extent, Carnap viewed science as a tool for 
social reform. 

The precise formulation of what came to be called the criterion of cognitive signifi- 
cance took three decades. (SeeHempel 1950 and Carnap 1956 and 1961.) In an impor- 
tant pair of papers, "Testability and Meaning," Carnap (1936, 193 7a) replaced the 
requirement of verification with that of confirmation; at this stage, he made no attempt 
to quantify the latter. Individual terms replace sentences as the units of meaning. 
Universal generalizations are no longer problematic; though they cannot be con- 
clusively verified, they can yet be confirmed. Moreover, in "Testability and Meaning," 
theoretical terms no longer require explicit definition from observational ones in order 
to acquire meaning; the connection between the two may be indirect through a system 
of implicit definitions. Carnap also provides an important pioneering discussion of 
disposition predicates. 

The syntactic phase 

Meanwhile, in 1 9 3 1 , Carnap had moved to Prague, where he held the Chair for Natural 
Philosophy at the German University until 1935 when, under the shadow of Hitler, he 
emigrated to the United States. Towards the end of his Vienna years, a subtle but impor- 
tant shift in Carnap's philosophical interests had taken place. This shift was from a 
predominant concern for the foundations of physics to that for the foundations of 
mathematics and logic, even though he remained emphatic that the latter were impor- 
tant only insofar as they were used in the empirical sciences, especially physics. 

In Vienna and before, following Frege and Russell, Carnap espoused logicism in its 
conventional sense, that is, as the doctrine that held that the concepts of mathematics 
were definable from those of logic and the theorems of mathematics were derivable 
from the principles of logic. In the aftermath of Godel's (1931) incompleteness theo- 
rems (see TARSKI, CHURCH, GODEL), however, Carnap abandoned this type of logicism and 
opted, instead, for the requirement that the concepts of mathematics and logic always 
have their customary, that is, everyday interpretation in all contexts. He also began to 
advocated a strong conventionalism regarding what constituted "logic." 



Besides the philosophical significance of Godel's results, what impressed Carnap 
most about that work was Godel's arithmetization of syntax. Downplaying the distinc- 
tion between an object language and its metalanguage, Carnap interpreted this proce- 
dure as enabling the representation of the syntax of a language within the language 
itself. At this point Carnap had not yet accepted the possibility of semantics even though 
he was aware of some of Tarski's work and had had some contact with the Polish school 
of logic. In this context, the representation of the syntax of a language within itself 
suggested to Carnap that all properties of a language could be studied within itself 
through a study of syntax. 

These positions were codified in Carnap's major work from this period, The Logical 
Syntax of Language (Carnap 193 7b). The English translation includes material that had 
to be omitted from the German original owing to a shortage of paper; the omitted 
material was separately published in German as papers (Carnap 1934a, 193 5). 
Conventionalism about logic was incorporated into the well-known Principle of 

It is not our business to set up prohibitions but to arrive at conventions [about what constitutes 
a logic] . ... In logic, there are no morals. Every one is at liberty to build up his own logic, i.e., 
his own form of language, as he wishes. All that is required is that, if he wishes to discuss 
it, he must state his method clearly, and give syntactic rules instead of philosophical argu- 
ments. (193 7b: 51-2; emphasis in the original) 

Logic, therefore, is nothing but the syntax of language. 

In Syntax, the Principle of Tolerance allows Carnap to navigate the ongoing disputes 
between logicism, formalism, and intuitionism/constructivism in the foundations of 
mathematics without abandoning any insight of interest from these schools. Carnap 
begins with a detailed study of the construction of two languages, I and II. The last few 
sections of Syntax also present a few results regarding the syntax of any language and 
also discuss the philosophical ramifications of the syntactic point of view. 7 

Language I, which Carnap calls "definite," is intended as a neutral core of all logi- 
cally interesting languages, neutral enough to satisfy the strictures of almost any intui- 
tionist or constructivist. It permits the definition of primitive recursive arithmetic and 
has bounded quantification (for all x up to some upper bound) but not much more. Its 
syntax is fully constructed formally. Language II, which is "indefinite" for Carnap, is 
richer. It includes Language I and has sufficient resources for the formulation of all of 
classical mathematics and is, therefore, non-constructive. Moreover, Carnap permits 
descriptive predicates in each language. Thus, the resources of Language II are strong 
enough to permit, in principle, the formulation of classical physics. The important point 
is that, because of the Principle of Tolerance, the choice between Languages I and II or, 
for that matter, any other syntactically specified language, is not based on factual 
considerations. If one wants to use mathematics to study physics in the customary 
way, Language II is preferable since, as yet, non-constructive mathematics remains 
necessary for physics. But the adoption of Language II, dictated by the pragmatic 
concern for doing physics, does not make Language I incorrect. This was Carnap's 
response to the foundational disputes of mathematics: by tolerance they are defined out 
of existence. 



The price paid if one adopts the Principle of Tolerance is a radical conventionalism 
about what constitutes logic. Conventionalism, already apparent in Carnap's admission 
of both a phenomenalist and a physicalist possible basis for construction in the 
Aufbau, and strongly present in the works on the foundations of physics in the 1920s, 
had now been extended in Syntax to logic. As a consequence, what might be 
considered to be the most important question in any mathematical or empirical context 
- the choice of language - became pragmatic. This trend of relegating troublesome 
questions to the realm of pragmatics almost by fiat, thereby excusing them from 
systematic philosophical exploration, became increasingly prevalent in Carnap's views 
as the years went on. 

Syntax contained four technical innovations in logic that are of significance: (1) a 
definition of analyticity that, as was later shown by S. C. Kleene, mimicked Tarski's defi- 
nition of truth for a formalized language; (2) Carnap constructed a proof, indepen- 
dently of Tarski, that truth cannot be defined as a syntactic predicate in any consistent 
formalized language; (3) a rule for infinite induction (in Language I) that later came 
to be called the omega rule; and (4), most important, a generalization of Godel's first 
incompleteness theorem that has come to be called the fixed-point lemma. With respect 
to (4), what Carnap proved is that, in a language strong enough to permit arithmeti- 
zation, for any syntactic predicate, one can construct a sentence that would be inter- 
preted as saying that it satisfies that predicate, ff the chosen predicate is unprovability, 
one gets Godel's result. 

Besides the Principle of Tolerance, the main philosophical contribution of Syntax 
was the thesis that philosophy consisted of the study of logical syntax. Giving a new 
twist to the Vienna Circle's claim that metaphysical claims were meaningless, Carnap 
argues and tries to show by example that sentences making metaphysical claims are all 
syntactically ill-formed. Moreover, since the arithmetization procedure shows that all 
the syntactic rules of a language can be formulated within the language, even the rules 
that determine what sentences are meaningless can be constructed within the lan- 
guage. All that is left for philosophy is a study of the logic of science. But, as Carnap 
puts it: "The logic of science (logical methodology) is nothing else than the syntax 
of the language of science. ... To share this view is to substitute logical syntax for 
philosophy" (193 7b: 7-8). The claims of Syntax are far more grandiose - and more 
flamboyant - than anything in the Aufbau. 


In the late 1930s Carnap abandoned the narrow syntacticism of Syntax and, under the 
influence of Tarski and the Polish school of logic, came to accept semantics. With this 
move, Carnap's work enters its final mature phase. For the first time, he accepted that 
the concept of truth can be given more than pragmatic content. Thereupon, he turned 
to the systematization of semantics with characteristic vigor, especially after his immi- 
gration to the US where he taught at the University of Chicago from 1936 to 1952. 
In his contribution to the International Encyclopedia of Unified Science, in 1939, on the 
foundations of logic and mathematics, the distinctions between syntactic, semantic, 
and pragmatic considerations regarding any language are first presented in their 
mature form. 



Introduction to Semantics, which followed in 1942, develops semantics systematically. 
In Syntax Carnap had distinguished between two types of transformations on sen- 
tences: those involving "the method of derivation" or "d-method"; and those involving 
the "method of consequence" or "c-method." Both of these were supposed to be syn- 
tactic but there is a critical distinction between them. The former allows only a finite 
number of elementary steps. The latter places no such restriction and is, therefore, more 
"indefinite." Terms defined using the d-method ("d-terms") include "derivable," 
"demonstrable," "refutable," "resoluble," and "irresoluble"; the corresponding "c- 
terms" are "consequence," "analytic," "contradictory," "L-determinate," and "syn- 
thetic." After the conversion to semantics Carnap proposed that the c-method 
essentially captured what semantics allowed; the c-terms referred to semantic concepts. 

Thus semantics involves a kind of formalization, though one that is dependent on 
stronger inference rules than the syntactical ones. In this sense, as Church (1956: 65) 
has perceptively pointed out, Carnap (and Tarski) reduce semantics to formal rules, that 
is, syntax. Thus emerges the interpretation of deductive logic that has since become the 
textbook version, so commonly accepted that is has become unnecessary to refer to 
Carnap when one uses it. For Carnap, the semantic move had an important philo- 
sophical consequence: philosophy was no longer to be replaced just by the syntax of 
the language of science; rather, it was to be replaced by the syntax and the semantics 
of the language of science. 

Carnap's most original - and influential - work in semantics is Meaning and Necessity 
(1947), where the basis for an intensional semantics was laid down. Largely following 
Frege, intensional concepts are distinguished from extensional ones. Semantical rules 
are introduced and the analytic/synthetic distinction is clarified by requiring that any 
definition of analyticity must satisfy the (meta-)criterion that analytic sentences 
follow from the semantical rules alone. By now Carnap had fully accepted that 
semantic concepts and methods are more fundamental than syntactic ones: the retreat 
from the flamboyance of Syntax was complete. The most important contribution of 
Meaning and Necessity was the reintroduction into logic, in the new intensional frame- 
work, of modal concepts that had been ignored since the pioneering work of Lewis 
( 1 9 1 8 ) . In the concluding chapter of his book Carnap introduced an operator for neces- 
sity, gave semantic rules for its use, and showed how other modal concepts such as 
possibility, impossibility, necessary implication, and necessary equivalence can be 
defined from this basis. 

By this point, Carnap had begun to restrict his analyses to exactly constructed lan- 
guages, implicitly abandoning even a distant hope that they would have any direct 
bearing on natural languages. The problem with the latter is that their ambiguities 
made them unsuited for the analysis of science which, ultimately, remained the moti- 
vation of all of Carnap's work. Nevertheless, Carnap's distinction between the analytic 
and the synthetic came under considerable criticism from many, including Quine 
(1951), primarily on the basis of considerations about natural languages. Though 
philosophical fashion has largely followed Quine on this point, at least until recently, 
Carnap was never overly impressed by this criticism (Stein 1992). The analytic/syn- 
thetic distinction continued to be fundamental to his views and, in a rejoinder to Quine, 
Carnap argued that nothing prevented empirical linguistics from exploring intensions 
and thereby discovering cases of synonymy and analyticity (Carnap 1955). 



Carnap's most systematic exposition of his final views on ontology is also from this 
period (1950a). A clear distinction is maintained between questions that are internal 
to a linguistic framework and questions that are external to it. The choice of a linguis- 
tic framework is to be based not on cognitive but on pragmatic considerations. The 
external question of "realism," which ostensibly refers to the "reality" of entities of a 
framework in some sense independent of it, rather than to their "reality" within it after 
the framework has been accepted, is rejected as non-cognitive. This appears to be an 
anti-"realist" position but is not in the sense that, within a framework, Carnap is tol- 
erant of the abstract entities that bother nominalists. The interesting question becomes 
the pragmatic one, that is, what frameworks are fruitful in which contexts, and 
Carnap's attitude towards the investigation of various alternative frameworks remains 
characteristically and consistently tolerant. 

Carnap continued to explore questions about the nature of theoretical concepts and 
to search for a criterion of cognitive significance, preoccupations of the logical empiri- 
cists that date back to the Vienna Circle. In 1956 he published a detailed exposition of 
his final views regarding the relation between the theoretical and observational parts 
of a scientific language (Carnap 1956). This paper emphasizes the methodological and 
pragmatic aspects of theoretical concepts. 

It also contains his most subtle, though not his last, attempt to explicate the notion 
of the cognitive significance of a term and thus establish clearly the boundary between 
scientific and nonscientific discourse. However, the criterion he formulates makes 
theoretical terms significant only with respect to a class of terms, a theoretical lan- 
guage, an observation language, correspondence rules between them, and a theory. 
Relativization to a theory is critical to avoiding the problems that beset earlier attempts 
to find such a criterion. Carnap proves several theorems that are designed to show that 
the criterion does capture the distinction between scientific and nonscientific discourse. 
This criterion was criticized by Roozeboom (1960) and Kaplan (1975) but these criti- 
cisms depend on modifying Carnap's original proposal in important ways. According 
to Kaplan, Carnap accepted his criticism though there is apparently no independent 
confirmation of that fact. However, Carnap (1961) did turn to a different formalism 
(Hilbert's e-operator) in his last attempt to formulate such a criterion and this may indi- 
cate dissatisfaction with the 1956 attempt. If so, it remains unclear why: that attempt 
did manage to avoid the technical problems associated with the earlier attempts of the 
logical empiricists. 

Inductive logic 

From 1941 onwards Carnap also began a systematic attempt to analyze the concepts 
of probability and to formulate an adequate inductive logic (a logic of confirmation), a 
project that would occupy him for the rest of his life. Carnap viewed this work as an 
extension of the semantical methods that he had been developing for the last decade. 
This underscores an interesting pattern in Carnap's intellectual development. Until the 
late 1930s Carnap only viewed syntactic categories as non-pragmatically specifiable; 
questions of truth and confirmation were viewed as pragmatic. His conversion to 
semantics saw the recovery of truth from the pragmatic to the semantic realm. Now, 
confirmation followed truth down the same pathway. 



In Logical Foundations of Probability (1950b), his first systematic analysis of prob- 
ability, Carnap distinguished between two concepts of probability: "statistical probabil- 
ity," which was the relevant concept to be used in empirical contexts and generally 
estimated from the relative frequencies of events, and "logical probability," which was 
to be used in contexts such as the confirmation of scientific hypotheses by empirical 
data. Though the latter concept, usually called the "logical interpretation" of probabil- 
ity went back to Keynes (1921), Carnap provides its first systematic explication. 

Logical probability is explicated from three different points of view (1950b: 164-8): 
(1) as a conditional probability c(h,e) which measures the degree of confirmation of a 
hypothesis h on the basis of evidence e (if c(h,e) = r, then r is determined by logical rela- 
tions between h and e); (2) as a rational degree of belief or fair betting quotient (if c(h,e) 
= r, then r is a fair bet on h if e correctly describes the total knowledge available to a 
bettor); and (3) as the limit of relative frequencies in some cases. According to Carnap, 
the first of these, which specifies a confirmation function ("ofunction"), is the concept 
that is most relevant to the problem of induction. In the formal development of the 
theory, probabilities are associated with sentences of a formalized language. 

In Foundations, Carnap believed that a unique measure c(h,e) of the degree of 
confirmation can be found and he even proposed one (namely, Laplace's rule of 
succession) though he could not prove its uniqueness satisfactorily. His general strat- 
egy was to augment the standard axioms of the probability calculus by a set of 
"conventions on adequacy" (1950b: 285), which turned out to be equivalent to 
assumptions about the rationality of degrees of belief that had independently been 
proposed by both Ramsey and de Finetti (Shimony 1992). In a later work, The 
Continuum of Inductive Methods (1952), using the conventions on adequacy and 
some plausible symmetry principles, Carnap managed to show that all acceptable c- 
functions could be parameterized by a single parameter, a real number, A e [0,°°]. The 
trouble remained that there is no intuitively appealing a priori strategy to restrict X to 
some preferably very small subset of [0,°°]. At one point, Carnap even speculated that 
it would have to be fixed empirically. Unfortunately, some higher-order induction would 
then be required to justify the procedure for its estimation and, potentially, this leads to 
infinite regress. 

Carnap spent 1952-4 at the Institute for Advanced Study at Princeton where he 
continued to work on inductive logic, often in collaboration with John Kemeny He also 
returned to the foundations of physics, apparently motivated by a desire to trace and 
explicate the relations between the physical concept of entropy and an abstract concept 
of entropy appropriate for inductive logic. His discussion with physicists proved to be 
disappointing and he did not publish his results. 8 

In 1954 Carnap moved to the University of California at Los Angeles to assume the 
chair that had become vacant with Reichenbach's death in 1953. There he continued 
to work primarily on inductive logic, often with several collaborators, over the next 
decade. There were significant modifications of his earlier attempts to formulate a sys- 
tematic inductive logic. 9 Obviously impressed by the earlier work of Ramsey and de 
Finetti, Carnap (1971b) returned to the second of his three 1950 explications of logical 
probability and emphasized the use of inductive logic in decision problems. 

More importantly, Carnap, in "A Basic System of Inductive Logic" (1971a, 1980) 
finally recognized that attributing probabilities to sentences was too restrictive. If a 



conceptual system uses real numbers and real-valued functions, no language can 
express all possible cases using only sentences or classes of sentences. Because of this, 
he now began to attribute probabilities to events or propositions (which are taken to be 
synonymous). This finally brought some concordance between his formal methods 
and those of mathematical statisticians interested in epistemological questions. 
Propositions are identified with sets of models; however, the fields of the sets are defined 
using the atomic propositions of a formalized language. Thus, though probabilities are 
defined as measures of sets, they still remain relativized to a particular formalized 
language. Because of this, and because the languages considered remain relatively 
simple (mostly monadic predicate languages) much of this work remains similar to the 
earlier attempts. 

By this point Carnap had abandoned the hope of finding a unique c-function. 
Instead, he distinguished between subjective and objective approaches in inductive 
logic. The former emphasizes individual freedom in the choice of necessary conven- 
tions; the latter emphasizes the existence of limitations. Though Carnap characteristi- 
cally claimed to keep an open mind about these two approaches, his emphasis was on 
finding rational a priori principles which would systematically limit the choice of c- 
functions. Carnap was still working on this project when he died on 14 September 
1970. He had not finished revising the last sections of the second part of the "Basic 
System," both parts of which were only published posthumously. 

Towards the end of his life, Carnap's concern for political and social justice had led 
him to become an active supporter of an African-American civil rights organization in 
Los Angeles. According to Stegmiiller (1972: lxvi), the "last photograph we have of 
Carnap shows him in the office of this organization, in conversation with various 
members. He was the only white in the discussion group." 

The legacy 

Some decades after Carnap's death it is easier to assess Carnap's legacy, and that of 
logical empiricism, than it was in the 1960s and 1970s when a new generation of ana- 
lytic philosophers and philosophers of science apparently felt that they had to reject 
that work altogether in order to be able to define their own philosophical agendas. This 
reaction can itself be taken as evidence of Carnap's seminal influence but, neverthe- 
less, it is fair to say that Carnap and logical empiricism fell into a period of neglect 
in the 1970s from which it only began to emerge in the late 1980s and early 1990s. 
Meanwhile it became commonplace among philosophers to assume that Carnap's 
projects had failed. 

Diagnoses of this failure have varied. For some it was a result of the logical empiri- 
cists' alleged inability to produce a technically acceptable criterion for cognitive signifi- 
cance. For others, it was because of Quine's dicta against the concept of analyticity and 
the analytic/synthetic distinction. Some took Popper's work to have superseded that of 
Carnap and the logical empiricists. Many viewed Kuhn's seminal work on scientific 
change to have shown that the project of inductive logic was misplaced (see KUHN); they, 
and others, generally regarded Carnap's attempt to explicate inductive logic to have 
been a failure. Finally, a new school of "scientific realists" attempted to escape Carnap's 
arguments against external realism. 



There can be little doubt that Carnap's project of founding inductive logic has 
faltered. He never claimed that he had gone beyond preliminary explorations of pos- 
sibilities and though there has been some work since, by and large, epistemologists of 
science have abandoned that project in favor of less restrictive formalisms, for instance, 
those associated with Bayesian or Fisherian statistics. But, with respect to every other 
case mentioned in the last paragraph, the situation is far less clear, ft has already been 
noted that Carnap's final criterion for cognitive significance does not suffer from any 
technical difficulty no matter what its other demerits may be. Quine's dicta against ana- 
lyticity no longer appear as persuasive as they once did (Stein 1992); Quine's prefer- 
ence for using natural - rather than formalized - language in the analysis of science 
has proved to be counterproductive; and his program of naturalizing epistemology is 
yet to live up to any initial promise that it ever might have had. Putnam's "internal 
realism" is based on and revives Carnap's views on ontology and Kuhn is perhaps now 
better regarded as having contributed significantly to the sociology rather than the epis- 
temology of science. 

However, to note that some of the traditionally fashionable objections to Carnap and 
logical empiricism cannot be sustained does not show that that work deserves a posi- 
tive assessment on its own. We are still left with the question: what, exactly, did Carnap 
contribute? The answer turns out to be surprisingly straightforward: the textbook 
picture of deductive logic that we have today is the one that Carnap produced in the 
early 1940s after he came to acknowledge the possibility of semantics. The fixed-point 
lemma has turned out to be an important minor contribution to logic. The reintroduc- 
tion of modal logic into philosophy opened up new vistas for Kripke and others in the 
1950s and 1960s (see KRIPKE). Carnap's views on ontology continue to influence 
philosophers today. Moreover, even though the project of inductive logic seems 
unsalvageable to most philosophers it is hard to deny that Carnap managed to clarify 
significantly the ways in which concepts of probability must be deployed in the 
empirical sciences and why the problem of inductive logic is so difficult. But, most of 
all, Carnap took philosophy to a new level of rigor and clarity, accompanied by an open- 
mindedness (codified in the Principle of Tolerance) that, unfortunately, has not been 
widely shared in analytic philosophy 10 


1 Biographical details are from Carnap 1963a. 

2 References to the Aufbau are to sections; this permits the simultaneous use of the German 
and English editions. 

3 An excellent discussion of Carnap's construction is to be found in Goodman 1951, ch. 5. 

4 Thus, if an elex is both c in color and t in temperature, c or t can be defined as classes of 
every elex having c or t respectively. However, to say that c or t is a quality would imply that 
an elex is analyzable into simpler constituents. Ouasi-analysis proceeds formally in this 
way (as if it is proper analysis) but only defines quasi-characteristics thus leaving each elex 

5 Goodman (1951) also provides a very lucid discussion of these problems. 

6 Some recent scholarship has questioned whether Carnap had any traditional epistemologi- 
cal concerns in the Aufbau. In particular, Friedman (e.g. 1 99 2 ) has championed the view that 
Carnap's concerns in that work are purely ontological: the Aufbau is not concerned with the 



question of the source or status of our knowledge of the external world; rather, it investigates 
the bases on which such a world may be constructed. (See, also, Richardson 1998. Both 
Friedman and Richardson - as well as Sauer (1985) and Haack (1977) long before them - 
emphasize the Kantian roots of the Aufbau.) If this reinterpretation is correct, then what 
exactly the Aufbau owes to Russell (and traditional empiricism) becomes uncertain. However, 
as Putnam (private communication) has pointed out, this reinterpretation goes too far: 
though the project of the Aufbau is not identical to that of Russell's external world program 
(for reasons including those that Friedman gives), there is sufficient congruence between the 
two projects for Carnap to have correctly believed that he was carrying out Russell's program. 
In particular, the formal constructions of the Aufbau are a necessary prerequisite for the 
development of the epistemology that Russell had in mind: one must be able to construct the 
world formally from a phenomenalist basis before one can suggest that this construction 
shows that the phenomena are the source of our knowledge of the world. Moreover, this rein- 
terpretation ignores the epistemological remarks scattered throughout the Aufbau itself, 
including Carnap's concern for the epistemic primacy of the basis he begins with. 

7 Sarkar (1992) attempts a comprehensible reconstruction of the notoriously difficult 
formalism of Syntax. 

8 These were edited and published by Abner Shimony (as Two Essays on Entropy, 19 77) after 
Carnap's death. 

9 See Carnap and Jeffrey 19 71 and Jeffrey 1980. An excellent introduction to this part of 
Carnap's work on inductive logic is Hilpinen 1975. 

10 For comments on earlier versions of this essay thanks are due to Justin Garson, Cory Juhl, 
Al Martinich, and Itai Sher. 


Works by Carnap 

1922: Der Raum, Berlin: von Reuther und Reichard. 

1923: "Uber die Aufgabe der Physik und die Anwendung des Grundsatzes der Einfachsteit," 

Kant-Studien 28, pp. 90-107. 
1924: "Dreidimensionalitat des Raumes und Kausalitat: Eine Untersuchung iiber den logischen 

Zusammenhang zweier Fiktionen," Annalen der Philosophie und philosophischen Kritik 4, 

pp. 105-30. 
1926: Physikalische Begriffsbildung, Karlsruhe: Braun. 
1934a: "Die Antinomien und die Unvollstandigkeit der Mathematik," Monatshefte fiir Mathematik 

und Physik 41, pp. 42-8. 
1934b: The Unity of Science, London: Kegan Paul, Trench, Trubner and Co. (Originally published 

193 5: "Ein Gultigskriterium fiir die Satze der klassischen Mathematik," Monatshefte fiir 

Mathematik und Physik 42, pp. 163-90. 
1936: "Testability and Meaning," Philosophy of Science 3, pp. 419-71. 
193 7a: "Testability and Meaning," Philosophy of Science 4, pp. 1-40. 
193 7b: The Logical Syntax of Language, London: Kegan Paul, Trench, Trubner and Co. (Originally 

Logische Syntax der Sprache, Vienna: Springer, 1934.) 
1939: Foundations of Logic and Mathematics, Chicago: University of Chicago Press. 
1947: Meaning andNecessity: A Study in Semantics and Modal Logic, Chicago: University of Chicago 

1950a: "Empiricism, Semantics, and Ontology," Revue Internationale de Philosophie 4, 

pp. 20-40. 



1950b: Logical Foundations of Probability, Chicago: University of Chicago Press. 

1952: The Continuum of Inductive Methods, Chicago: University of Chicago Press. 

1955: "Meaning and Synonymy in Natural Languages," Philosophical Studies 7, pp. 33-47. 

1956: "The Methodological Character of Theoretical Concepts," in The Foundations of Science 

and the Concepts of Psychology and Psychoanalysis, ed. H. Feigl and M. Scriven, Minneapolis: 

University of Minnesota Press, pp. 38-76. 
1961: "On the Use of Hilbert's e-Operator in Scientific Theories," in Essays on the Foundations of 

Mathematics, ed. Y. Bar-Hillel, E. I. J. Poznanski, M. 0. Rabin, and A. Robinson, Jerusalem: 

Magnes Press, pp. 156-64. 
1963a: "Intellectual Autobiography," in The Philosophy of Rudolf Carnap, ed. P. A. Schilpp, La 

Salle, IL: Open Court, pp. 3-84. 
1963b: "Replies and Systematic Expositions," in The Philosophy of Rudolf Carnap, ed. P. A. Schilpp, 

La Salle, IL: Open Court, pp. 859-1013. 
1967: The Logical Structure of the World and Pseudoproblems in Philosophy, Berkeley: University of 

California Press. (Originally published 1928.) 
1971a: "A Basic System of Inductive Logic, Part I," in Studies in Inductive Logic and Probability , 

ed. R. Carnap and R. C. Jeffrey, vol. 1, Berkeley: University of California Press, pp. 33-165. 
1971b: "Inductive Logic and Rational Decisions," in Studies in Inductive Logic and Probability , 

ed. R. Carnap and R. C. Jeffrey, vol. 1, Berkeley: University of California Press, pp. 5-31. 
1977: Two Essays on Entropy, Berkeley: University of California Press. 
1980: 'A Basic System of Inductive Logic, Part II," in Studies in Inductive Logic and Probability , 

ed. R. C. Jeffrey, vol. 2, Berkeley: University of California Press, pp. 8-155. 

1971 (with Jeffrey, R. C.) (eds.): Studies in Inductive Logic and Probability, vol. 1, Berkeley: 
University of California Press. 

Works by other authors 

Ayer, A. J. (1936) Language, Truth andLogic, London: Gollancz. 

Church, A. (1956) Introduction to Mathematical Logic, vol. 1, Princeton, NJ: Princeton University 

Friedman, M. (1992) "Epistemology in the Aufbau," Synthese 93, pp. 191-23 7. 
Godel, K. ( 1 9 3 1 ) "Uber formal unentscheidbare Satze der Principia Mathematica und verwandter 

Systeme 1," Monatshefte fur Mathematik undPhysik 38, pp. 173-98. 
Goodman, N. (1951) The Structure of Experience, Cambridge, MA: Harvard University Press. 
Haack, S. (19 77) "Carnap's Aufbau: Some Kantian Reflections," Ratio 19, pp. 170-5. 
Hempel, C. G. (1950) "Problems and Changes in the Empiricist Criterion of Meaning," Revue 

Internationale de Philosophic 11, pp. 41-63. 
Hilpinen, R. (19 75) "Carnap's New System of Inductive Logic," in Rudolf Carnap, Logical 

Empiricist: Materials and Perspectives, ed. J. Hintikka, Dordrecht: Reidel, pp. 333-59. 
Jeffrey, R. C. (ed.) (1980) Studies in Inductive Logic and Probability, vol. 2, Berkeley: University of 

California Press. 
Kaplan, D. (1975) "Significance and Analyticity: A Comment on Some Recent Proposals of 

Carnap," in Rudolf Carnap, Logical Empiricist: Materials and Perspectives, ed. J. Hintikka, 

Dordrecht: Reidel, pp. 87-94. 
Keynes, J. M. (1921) A Treatise on Probability, London: Macmillan. 
Lewis, C. I. (1918) A Survey of Symbolic Logic, Berkeley: University of California Press. 
Quine, W. V (1951) "Two Dogmas of Empiricism," Philosophical Review 60, pp. 20-43. 
Richardson, A. (1998) Carnap's Construction of the World, Cambridge: Cambridge University 




Roozeboom, W. (1960) "A Note on Carnap's Meaning Criterion," Philosophical Studies 11, pp. 33-8. 
Sarkar, S. (1992) "'The Boundless Ocean of Unlimited Possibilities': Logic in Carnap's Logical 

Syntax of Language," Synthese 93, pp. 191-23 7. 
Sauer, W. (1985) "Carnap's 'Aufbau' in Kantianischer Sicht," Grazer Philosophische Studien 23, 

pp. 19-3 5. 
Shimony, A. (1992) "On Carnap: Reflections of a Metaphysical Student," Synthese 93, 

pp. 261-74. 
Stegmuller, W. (19 72) "Homage to Rudolf Carnap," in PSA 1970, ed. R. C. Buck and R. S. Cohen, 

Dordrecht: Reidel, pp. lii-lxvi. 
Stein, H. (1992) "Was Carnap Entirely Wrong, After All?," Synthese 93, pp. 2 75-95. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

Karl Popper (1902-1994) 


Born in Vienna, Karl Popper studied at the University of Vienna from 1918 to 1922, 
after which he became apprenticed to a master cabinetmaker, Adalbert Posch. In his 
intellectual autobiography, Popper reported that he learned more about epistemology 
from Posch than from any other of his teachers. In 1925 he enrolled in the City of 
Vienna's new Pedagogic Institute to work on the psychology of thought and discovery. 
However, his interests turned to methodology and in 1928 he obtained his doctorate 
for a thesis on methodological problems in psychology. 

While teaching mathematics and physics in a secondary school he wrote his Logik 
der Forschung, which was published in 1934, appearing in an English translation in 
1959 as The Logic of Scientific Discovery. In 1 9 3 7 he went to New Zealand as a lecturer 
in philosophy at Canterbury University College. While there he wrote his influential 
works The Poverty of Historicism and The Open Society and its Enemies. Appointed Reader 
and subsequently Professor in Logic and Scientific Method at the London School of 
Economics in 1946 he remained there for the balance of his academic career. Until 
his death in 1994 he continued to publish prolifically. 

The distinctive feature of Popper's philosophy of science is his attitude to induction. 
Like Hume he held that no inductive inference is ever rationally justified. Finding that 
1 million randomly selected samples of sodium burn with a yellow flame provides no 
reason at all, according to Hume and Popper, for thinking that all pieces of sodium will 
burn with a yellow flame. What we would normally count as evidence for such a 
hypothesis does not even give fallible grounds for thinking it is more probably true than 
false. Inductive arguments, arguments in which the premises do not entail the conclu- 
sion but purport to support it, simply have no rational force. Consequently Popper 
sought to rely entirely on deductive argumentation. While we can never have the least 
positive reason for thinking that a hypothesis is true or probably true, we can use a 
deductive argument to show that it is false. For given that we have observed one black 
swan we can deductively infer that it is false that all swans are white. This is the crux 
of Popper's philosophy of science. It is only the rejection of beliefs or hypotheses that 
can have the sanction of reason (but see below). 

Hume never sought to persuade us to abandon induction. For him, it is part of 
our nature to proceed inductively. Custom and habit carry us forward where reason 
fails. To put the point anachronistically, for Hume we are "hard-wired" to induct. It is 



simply a bemusing feature of the human condition that our inductive procedures do 
not have the sanction of reason. But for Popper, on the other hand, we do not or should 
not proceed inductively. And he claims that good scientists never do so. The Popperian 
scientist, equipped with a fertile imagination, simply makes a bold conjecture and the 
bolder the better. He then seeks to refute that conjecture by observation and experi- 
mentation. If a contrary instance is found the conjecture is falsified and hence rejected. 
In which case the scientist starts again with a new conjecture. If a conjecture is not 
falsified in a test, it has been "corroborated." Corroboration, as defined by Popper, does 
not provide any reason for thinking that the hypothesis has any likelihood of holding 
in the future. It is simply a report that it has not yet failed. Critics have wondered why, 
in this case, we should trust even a highly corroborated hypothesis. Clearly we do so 
when, for example, we trust our fate to airplanes designed on the basis of aeronautical 
theories. The answer for Popper is that we have no reason at all to do so. We proceed 
on blind faith! Critics also object that rejecting a hypothesis in the face of a contrary 
instance is itself a disguised form of induction. For in so doing we are assuming that 
the future will be like the past: what failed on Monday will also fail on Tuesday. 

In utterly rejecting anything other than deductive justification, Popper committed 
himself to a very strong form of fallibilism, according to which not only can we not 
have certain knowledge in science or in everyday life, we can have no positive reasons, 
however weak, for holding that particular beliefs in science or in everyday life are even 
more likely to be true than false. Some readers of Popper may well have failed to see the 
extreme consequences of his fallibilism. For it applies also to the beliefs we form about 
what we observe. Consequently, it follows that we can have no rational grounds for 
claiming to have discovered that a hypothesis had been falsified. For we can have no 
more reason for rejecting a hypothesis than we have for our belief that we have observed 
a counterexample. That being so, Popper's fallibilism amounts to an extreme form of 
skepticism. We can have no reasons for thinking that any empirical proposition is true; 
nor can we have any reasons for thinking that it is false. Much of what Popper wrote 
has plausibility only if we set aside this extreme consequence of his position and this 
I will do in much of what follows. 

Falsification provided Popper with his criterion for the demarcation of science from 
non-science or pseudo-science. He described this criterion as the very center of his phi- 
losophy of science. Impressed positively by the success of Einstein and negatively by 
what he took to be the failure of Freud and Marx, he looked for the hallmark of the sci- 
entific and thought he had found it in falsification. A theory is scientific just in case it 
makes predictions that could in principle be observed not to obtain. If they do not obtain 
the theory is refuted. According to Popper psychoanalysis ruled nothing out and hence 
could not be falsified and was not scientific. He held that Marxism was originally falsi- 
fiable. However, in the face of negative instances, Marxism was revised so as to become 
immune from refutation. In Popper's terminology any such unfalsifiable theories are 
metaphysical rather than scientific. Unfortunately, much of what we count as science 
turns out not to be falsifiable. A theory, such as quantum mechanics, which makes only 
probabilistic predictions, is not falsifiable. Consider the hypothesis that the probability 
that this coin will land heads on the next toss is p. The coin may land tails any number 
of times without falsifying that hypothesis! Popper sought to avoid this difficulty by 
adopting a methodological rule that would reject this hypothesis if after some number 



N of trials the results diverged significantly from p. Critics have not been satisfied that 
there is any reasonable way of fixing an appropriate value for N without the use of 
inductive argumentation. 

Popper did come to appreciate that there are metaphysical elements in good scien- 
tific theories and that metaphysical theories such as Darwinism had important bene- 
ficial influences on scientific development. Having recognized this Popper shifted some- 
what and sought to evaluate metaphysical theories as well. This is to be done by con- 
sidering whether the theory solves problems, whether its purported solutions can be 
examined critically and whether it solves the problems better than rival theories. This 
in turn generates a more general demarcation criterion for distinguishing between 
what he referred to as criticizable versus non-criticizable theories, ft is no doubt a step 
forward to consider the merits of theories in general without special regard to whether 
they are scientific or not. But it means abandoning what once had pride of place within 
his philosophy of science: a hallmark of the scientific. Even Freud and Marx meet the 
condition of being criticizable theories, as Popper's own writings make manifest. 

The method of science is to propose bold theories, the bolder the better. The scien- 
tist then seeks to refute them. Devoting oneself merely to finding out that theories are 
false does not seem a very edifying vocation, ff there were only a finite number of theo- 
ries in any branch of science, one could take comfort in the fact that with each rejec- 
tion, the probability of the next theory selected being true would increase. But unhap- 
pily there are an infinite number of rival theories. A Popperian scientist expects that 
even his most cherished theory will eventually be falsified, ff all he can ever find is that 
a theory is false, what positive gloss can he put on his scientific endeavors? For Popper 
the scientist hopes to have theories - false theories - that are ever better approxima- 
tions to the truth. These are theories with, in his words, "increasing truthlikeness or 
verisimilitude." The move from Newton to Einstein was progressive, because while 
Newton said some true things about the world and some false things, Einstein said more 
true things and fewer false things. We can picture this as Newton getting a certain per- 
centage of his claims right and Einstein scoring a higher percentage. Some future sci- 
entist can hope for a higher score yet. The idea that the aim of science is not truth per 
se but ever more approximately true theories has attracted adherents including many 
who reject Popper's account of scientific method. Unfortunately Popper's own techni- 
cal definition of verisimilitude proved unsatisfactory, ft turned out that on his defini- 
tion all theories other than true ones have the same degree of truthlikeness. Popper's 
approach has inspired much further work on this notion but at present no satisfactory 
explication of truthlikeness has been forthcoming. 

Popper is convinced that science is generating ever more truthlikeful theories. But 
as someone who avoids all inductive argumentation he has to regard this belief as irra- 
tional. Intuitively we might argue that the fact that Einstein passes more tests than 
Newton gives us reasonable grounds for thinking that Einstein's theory is more truth- 
likeful than Newton's. But this is an inductive argument. The conclusion is reasonable 
only if we assume that the area of the universe we have explored to date is a represen- 
tative sample of the entire universe. Perhaps it is just a local peculiarity that Einstein 
fares better than Newton. 

Popper himself is tempted by such arguments. At one point he claimed that we could 
argue for the greater truthlikefulness of Einstein over Newton on the grounds that it 



would be "a highly improbable coincidence if a theory like Einstein's could correctly 
predict very precise measurements not predicted by its predecessors unless there is 
'some truth' in it" (Schlipp 1974: 1192-3). This argument has many adherents par- 
ticularly among those who advocate realism about scientific theories. But it is a form of 
induction known as inference to the best explanation. We are invited to infer that 
Einstein's theory has more truth in it than Newton's theory on the grounds that that 
assumption provides the best explanation of the greater predictive success of Einstein's 
theory. Respectable as this argument may be, it is not open to one who rejects all but 
deductive argumentation. One might feel that in all fairness we should allow Popper 
just this one little inductive move. He himself concedes in the passage quoted above that 
"there may be a 'whiff of inductivismhere." But if an inductive move is legitimate here, 
why not elsewhere as well? Once we allow induction a role, Popper loses claim to our 
attention. For what made his philosophy of science unique and thereby interesting was 
its explicit and total rejection of induction. But without induction, his belief in scien- 
tific progress is irrational. 

Philosophers of science are divided on many issues. But they are almost unanimous 
in rejecting a Popperian account of science. Whether or not we have a satisfactory 
answer to Hume's skepticism about induction, it takes courage to deny that scientists 
proceed inductively. Scientists make an inductive move when they conclude that there 
is some probable truth in the theory of the electron on the grounds that that theory 
explains why televisions work. Even the scientist who concludes more modestly that 
there are at least good reasons to think that the theory of the electron will give suc- 
cessful observational predictions in the future is assuming the legitimacy of induction. 
Popper's falsificationist theory of science is itself falsified by scientific practice. His grand 
experiment in offering a non-inductivist theory of science serves only to heighten our 
appreciation of the deep-seated commitment to induction. 

Popper argued for a number of philosophical positions quite independent of his fal- 
sificationism. For instance, he was a passionate defender of the freedom of the will. He 
argued against the determinist's thesis that the future is fixed by the past states of the 
universe together with the laws of nature. His strategy was to seek to show that not all 
future events involving human agents can be scientifically predicted. However, critics 
have been unable to see how an inability even in principle to predict some future human 
actions shows that those actions may not be determined nonetheless. More is needed 
to establish freedom than an inability to predict. In addition he has urged a controver- 
sial three-part ontology. He posits a world of physical objects ("world 1"), a world of 
subjective experiences ("world 2"), and a world of the "objective contents of thought" 
("world 3"). "World 3" is reminiscent of the world of abstract Platonic objects that some 
philosophers have felt driven to postulate. But what is perhaps quite unique to Popper 
is the thought that this world, initially created by us, takes on an autonomy whereby 
it acts in a quasi-causal way on the objects of "world 2" and even of "world 1." Few 
philosophers have been willing to follow him in this lavish postulation of an unfalsifi- 
able theory of causally active abstract objects. 

History is likely to remember Popper more as a cultural figure than a philosopher in 
the narrow Anglo-Saxon sense of the term. For through his The Poverty of Historicism 
(1944) and his The Open Society and its Enemies (1945), he may well have had 
more influence in the social and cultural spheres than any other twentieth-century 



philosopher. His initial aim was to provide an intellectually decisive critique of 
Marx and Marxism. The resulting polemic became a hugely influential attack on 
totalitarianism in general and a glimpse of an inspiring if only vaguely sketched 
Utopia: the Open Society. 

Popper saw both fascism and Communism as resting on a pernicious historicism, a 
vision of history moving inevitably to some fixed final destination. Popper's historicist 
thinks he can detect by intuitive observation historical trends that he mistakenly takes 
to be iron laws and not mere trends that can be reversed. Popper characterizes Marxism, 
his paradigm of historicism, as bad historical prophecy combined with the injunction 
"Help to bring about the inevitable!" (1976: 3 5). Even granting Popper's assumption 
about the negative role that a belief in historicism has played and even granting the 
cogency of his arguments against it, his position does not really address the general 
problem of totalitarianism. For unfortunately there are totalitarian regimes that have 
come about as a result of forces other than belief in historicism. 

Popper, in contrast to the historicists' policy of waiting for the inevitable, grand- 
scale, social change, advocates "piecemeal social engineering." We should make small 
experimental adjustments in our social institutions; this he illustrates with such exam- 
ples as the introduction of a new sales tax. We then observe the results of the test; find 
out our errors and learn from our mistakes. For Popper this is explicitly an extension 
to the sphere of politics of the scientific method, although ironically the boldness that 
characterized the ideal scientist's conjectures is to be replaced by cautious small-scale 
conjectures. The bad moves are exposed; new ones are tried in their place. Our social 
engineer is urged to undertake a "systematic fight against definite wrongs, against con- 
crete forms of injustice or exploitation, and avoidable suffering such as poverty or 
unemployment" (1957: 91). 

In the social and political sphere, the notion of scientific rationality or reasonable- 
ness is liberalized to give a wider notion of rationality or reasonableness as openness to 
criticism: an attitude of readiness to listen to critical arguments and to learn from expe- 
rience. A commitment to rationality in this sense is a necessary core of Popper's lightly 
sketched vision of an open society. An open society is a democratic one that promotes 
criticism and diversity without repression or irreconcilable social divisions, avoids vio- 
lence, and encourages toleration. Critical public discussion with the participation of all 
is the means whereby those in an open society seek to arrive at a consensus on social 
and political issues. While the details are slight, Popper makes clear his touchingly naive 
view that the "free world" is a reasonable approximation to his ideal of an open society, 
having "very nearly, if not completely, succeeded in abolishing the greatest evils which 
have hitherto beset the social life of man" (1963: 3 70). 

The core of the idea of an open society is supposed to lie in the recognition of our 
fallibility. Popper's approach is reminiscent of Mill, who argued in On Liberty that we 
have an interest in promoting diversity of opinion. Beliefs may be incorrect. If we are 
wrong, our best hope in correcting ourselves lies in critical discussion with those who 
disagree. And if we are in fact correct, the critical discussion will bring out more clearly 
the contents of and grounds for our beliefs. But it is hard to see how any argument of 
this character can give us Popper's full conception of an open society. For instance, an 
interest in critical discussion is logically compatible with a majority maintaining a class 
of slaves whose role in life was to provide us with critical comment! Attractive as the 



nexus of values embodied in the open society are, they do not seem derivable just from 
the refutation of historicism together with the recognition of our fallibility. While 
society would no doubt benefit from more critical dialogue, it is arguable that the sta- 
bility of social institutions requires that not everything is open to debate at all times. 

There are serious tensions between the method of piecemeal social engineering and 
the advocacy of an open society. On the one hand, this method may only be viable for 
those who have already established an open society. The transitions in Albania and 
Romania, for example, from a closed to an open society were not achieved by piecemeal 
social engineering and probably could not have been achieved by anything other than 
revolution. On the other hand, it is not clear that all problems facing open societies can 
be solved through piecemeal social engineering. Globalization presents problems the 
resolution of which will require international regulatory mechanisms and these will 
not come about through piecemeal social engineering. 

For all the deficiencies in argumentation and conception, Popper's rhetoric has been 
much used and used to largely beneficial effect. The Poverty of Historicism and The Open 
Society were texts widely read in samizdat under Communism and certainly had an 
inspirational role for dissidents seeking to open their societies. At the same time these 
texts were widely used by western European social democratic parties of the left to dis- 
tance themselves from the Communist Parties of Europe. Progress to social equality was 
to be achieved by piecemeal social engineering, not revolutionary change. And in China 
in the period just before Tiananmen Square the liberal wing of the Communist Party 
used Popper explicitly in their analysis of the mistakes of the Cultural Revolution and 
in advocating open, critical discussion of social issues. 

Popper is no doubt right in encouraging more critical discussion of social and 
political issues. Unfortunately his own philosophical system, with which he seeks 
to underpin this encouragement, is limited. His utter rejection of induction, his 
fallibilism, severely restricts the scope of rational criticism. Given that only deductive 
argumentation has rational force, the only intellectually justified criticism that can be 
made of any position in science or in society is that the position is logically incoherent. 
But such a limited form of criticism is unlikely greatly to assist in the solution of the 
social ills that concerned Popper. 


Works by Popper 

1945: The Open Society and its Enemies, London: Routledge and Kegan Paul. 

1957: The Poverty of Historicism, London: Routledge and Kegan Paul. 

1959: The Logic of Scientific Discovery, London: Hutchinson. 

1963: Conjectures and Refutations, London: Routledge and Kegan Paul. 

1972: Objective Knowledge, Oxford: Oxford University Press. 

1976: Unended Quest: An Intellectual Autobiography, London: Collins. 

Works by other authors 

Newton-Smith, W. H. ( 1 9 8 1 ) The Rationality of Science, London: Routledge and Kegan Paul, ch. III. 
O'Hear, A. (1980) Karl Popper, London: Routledge and Kegan Paul. 



O'Hear, A. (ed.) (1995) Karl Popper: Philosophy and Problems, Cambridge: Cambridge University 

Raphael, F. (1998) Popper, London: Phoenix. 
Schilpp, P. A. (ed.) (1974) The Philosophy of Karl Popper, La Salle, IL: Open Court. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

Gilbert Ryle (1900-1976) 


Gilbert Ryle and his junior colleague, J. L. Austin, were the leading figures of post- World 
War II Oxford philosophy. Though their aims and methods were different (see below), 
both are correctly characterized as "ordinary language philosophers." Unlike Austin, 
who published only seven papers in his lifetime, Ryle was a prolific writer. Much of what 
we know about his personal life derives from self-references in his numerous biographi- 
cal sketches and reviews, and especially from his autobiography. In these various essays 
he describes his interactions with, and assessments of, the foremost philosophers of the 
time, among them Wittgenstein, Moore, Collingwood, Carnap, Prichard, H. H. Price, 
and Austin. His autobiography is to be found in Ryle, edited by 0. P. Wood and G. Pitcher 
(1970). Although it is only fifteen pages long, it is wittily self-deprecating, devastating 
in its depiction of the state of philosophy in Oxford in the 1920s and 1930s, packed 
with information, and instructive with respect to his philosophical development. About 
Oxford philosophy he says: 

During my time as an undergraduate and during my first years as a teacher, the philo- 
sophical kettle in Oxford was barely lukewarm. I think that it would have been stone cold 
but for Prichard, who did bring into his chosen and rather narrow arenas vehemence, 
tenacity, unceremoniousness, and a perverse consistency that made our hackles rise as 
nothing else at that time did. The Bradleians were not yet extinct, but they did not come 
out into the open. I cannot recollect hearing one referring mention of the Absolute. The 
Cook Wilsonians were hankering to gainsay the Bradleians and the Croceans, but were 
given few openings. Pragmatism was still represented by F. C. S. Schiller, but as his taste- 
less jocosities beat vainly against the snubbing primnesses of his colleagues, even this puny 
spark was effectually quenched . . . Soon Oxford's hermetically conserved atmosphere 
began to smell stuffy even to ourselves. 

About himself he states that in his mid-twenties he decided that philosophy essen- 
tially involves argumentation, and therefore that "the theory and technology" of rea- 
soning needed to be studied by any would-be philosopher. Since nothing of that sort 
was available in Oxford he "went all Cambridge," and seriously began to study Russell; 
but, as he frankly admits, with marginal qualifications: 

Having no mathematical ability, equipment or interest, I did not make myself even com- 
petent in the algebra of logic; nor did the problem of the foundations of mathematics 



become a question that burned in my belly. My interest was in the theory of Meanings - 
horrid substantive! - and quite soon, I am glad to say in the theory of its senior partner, 
Nonsense. I laboured upon the doublets - Sense and Reference, Intension and Extension, 
Concept and Object, Propositions and Constituents, Objectives and Objects, Facts and 
Things, Formal Concepts and Real Concepts, Proper Names and Descriptions, and Subjects 
and Predicates. It was in Russell's Principles of Mathematics and not in his Principia 
Mathematica, in his Meinong articles and his "On Denoting," that I found the pack-ice of 
logical theory cracking. It was up these cracks that Wittgenstein steered his Tractatus. 

His interests in the theories of meaning and reference were to dominate the remain- 
der of his career, and differentiate his version of ordinary language philosophy from 
Austin's. Austin's main concerns were in the utterances that constitute promises, 
warnings, recommendations, admonishments, counsels, and commands, - i.e. in so- 
called "speech acts" (see AUSTIN) - whereas Ryle saw his task as that of distinguishing 
locutions that make sense from those that do not. In a succinct passage Ryle explains 
the difference between his task and Austin's. 

An examiner might pose two questions: 

(1) Why cannot a traveller reach London gradually? 

(2) Why is "I warn you ..." the beginning of a warning, but "I insult you" not the 
beginning of an insult? 

On six days out of seven Question 1 would be Ryle's favourite; Question 2, Austin's. Each 
of us would think - wrongly - that there is not much real meat in the unfavoured 
question. But their meats are of such entirely disparate kinds that the epithet "linguistic" 
would apply in totally different ways (1) to the answer-sketch, "Adverbs like 'gradually' 
won't go with verbs like 'reach' for the following reason ..."; (2) to the answer-sketch "To 
insult is to say to someone else pejorative things with such and such an intention, while 
to warn is to say ..." Anti-nonsense rules govern impartially sayings of all types. "Reach 
gradually" will not do in questions, commands, counsels, requests, warnings, complaints, 
promises, insults, or apologies, any more than it will do in statements. Epimenides can 
tease us in any grammatical mood. To an enquiry into categorial requirements, references 
to differences of saying-type are irrelevant; to an enquiry into differences between saying- 
types, references to category-requirements are irrelevant. Infelicities and absurdities are 
not even congeners. 

As Ryle points out these different approaches were not in competition, but rather 
represented two parallel paths that "informal philosophy" could legitimately take in 
dealing with philosophical problems. Among those who emphasized the sense/non- 
sense distinction were Wittgenstein, Moore, J. T. Wisdom, 0. K. Bouwsma, and Norman 
Malcolm. Austin's focus on speech acts was later to influence the work of Paul Grice, 
Zeno Vendler, John Searle, and A. P. Martinich. And, of course, there are many philoso- 
phers, including Ryle and Austin and some of those just mentioned, in which both 
approaches play concurrent roles. 

In the twenty years between 1927 and 1947, Ryle had published more than thirty 
articles, reviews, and critical notices, but no books. His first venture into this larger 
format was The Concept of Mind (1949). Apart from collections of his essays he was to 
publish only two other books in his lifetime, Dilemmas in 1954, and Plato's Progress in 
1966. In the former book, Ryle discusses six tensions (dilemmas) that are not counter- 



vailing formal theories but rather opposing "platitudes." Each is an analogue of a clas- 
sical philosophical perplexity, such as the free will problem. Thus, "In card games and 
at the roulette-table it is easy to subside into the frame of mind of fancying that our 
fortunes are in some way prearranged, well though we know that it is silly to fancy 
this." Ryle shows by a subtle, piecemeal analysis of the linguistic idioms in which the 
opposing platitudes are framed how the apparent dilemma is factitious and can be dis- 
solved. Plato's Progress is an entirely different kind of book. It is a historical analysis in 
which Ryle tries to give a different portrayal of Plato's career. It is a provocative trea- 
tise that questions the common view that Aristotle was Plato's pupil, and that gives new 
datings to the Platonic dialogues. 

Though these monographs are exciting pieces, and well worth serious study, they do 
not match the power and depth of the Concept of Mind. It has two aspects: a negative, 
deflationary one and a positive, constructive one. The two approaches are tied together 
by an attack on a certain picture of the human mind and its relationship to the human 
body. Ryle gives different names to this picture: He calls it the "Official Doctrine," the 
"Cartesian Model," "Descartes' Myth," the "Ghost in the Machine," and the "Para- 
Mechanical Hypothesis." The negative attack is to show that this picture is incoherent; 
the positive contribution is to give an accurate account (not a picture) of the relation- 
ship between mind and body. The positive account is detailed. It deals with the entire 
range of the mental: the will, knowing, emotions, dispositions and occurrences, self- 
knowledge, sensation, observation, imagination, and the intellect. The book is thus a 
treasure-house of detailed descriptions of all the major features of mentation. 

What is the Official Doctrine he is out to destroy? This doctrine, he contends, is given 
its canonical formulation by Descartes, but its antecedents are much older. It is widely 
accepted by philosophers, psychologists, religious teachers, and many ordinary per- 
sons. It holds that every human being is both a mind and a body that are ordinarily 
harnessed together, but that after the death of the body the mind may continue to exist 
and function. Human bodies are in space and are subject to the mechanical laws of 
physics, chemistry, and biology. The body is a public object and can be inspected 
by external observers. But minds are immaterial, and are not in space, nor are their 
operations subject to mechanical laws. The mind is an entity, to be sure, but an imma- 
terial and invisible one that inhabits a mechanical body. This is why Ryle calls it the 
"ghost in the machine." It is res cogitans in Descartes' parlance. It is the thing that 
thinks, deliberates, decides, wills, and opines. Each mind is private, i.e. only each person 
can take direct cognizance of the states and processes of his or her own mind. 

A person thus lives through two collateral histories; one consisting of what happens 
to his body the other to what happens within his mind. The first is public, the second 
private. The Cartesian picture thus depends on the internal/external distinction. This 
leads to the problem of how the mind influences bodily action. Since the mind is con- 
strued as nonphysical and nonspatial how does one's act of will, say, lead to a move- 
ment of one's legs, i.e. to the sort of thing called walking, for instance? Moreover, how 
are we to account for the knowledge we presume we have of the minds of others? If the 
Cartesian model is correct, observers cannot know what is taking place in the mind of 
another, since they are in principle cut off from any sort of direct cognitive awareness 
of that person's mental states or processes. The only direct knowledge any human has 
is of his or her own mental functions. 



As plausible as this view may seem, it is absurd according to Ryle. It is one big mistake 
and a mistake of a special kind that he calls a "category mistake." To illustrate what he 
means by a "category mistake," Ryle offers three examples. Here is an abbreviated 
version of the first of these: 

A foreigner visiting Oxford or Cambridge for the first time is shown a number of colleges, 
libraries, playing fields, museums, scientific departments and administrative offices. He 
then asks "But where is the University? I have seen where the members of the Colleges live, 
where the Registrar works, where the scientists experiment and the rest. But I have not yet 
seen the University in which reside and work the members of your University." . . . His 
mistake lay in his innocent assumption that it was correct to speak of Christ Church, the 
Bodleian Library, the Ashmolean Museum and the University, to speak, that is, as if "the 
University" stood for an extra member of the class of which these other units are members. 
He was mistakenly allocating the University to the same category as that to which the other 
institutions belong. (1949: 16) 

Ryle's point is that this sort of mistake is made by people who do not know how to 
employ the concept of a university. That is, their puzzle arises from an inability correctly 
to use certain items in the English vocabulary. According to Ryle, the Official Doctrine 
arises from a category mistake analogous to the preceding, ft assumes that minds 
belong to the same category as bodies in the sense that both are rigidly governed by 
deterministic laws. The human body works according to mechanical principles: the 
heart is a pump, the veins are pipes, and the flow of blood is determined by the pres- 
sures that are described in fluid mechanics. The system is thus an assemblage of inter- 
acting parts that consist of fluids, solids, and electrical forces, all of which operate 
according to the laws of mechanics. All these forces usually work to some desired end, 
such as moving blood from one part of the body to another. 

Minds also work in analogous ways. When I am hungry, a mental state, a desire, acts 
on my body and initiates those movements of hands and fingers that allow me to pick 
up and transfer food to my mouth. Accordingly minds must be governed by determin- 
istic laws. But minds are nonmaterial. They are not composed of solids, fluids, and 
electrical forces. So their laws, though deterministic, are non-mechanical. These Ryle 
calls "para-mechanical." The Official Doctrine invokes them as the analogues of the 
mechanical laws that govern the behavior of physical entities. But the concept of a 
para-mechanical law is absurd. There are no such things as immaterial levers, valves, 
and pumps. Valves, levers, and pumps are solid entities that operate to effect physical 
movements. To invoke the immaterial analogues of such entities to explain mental 
activity is thus to make a category mistake, i.e. to apply the concepts of mechanical 
forces and laws to a domain where they have no grip. The mistake arises because 
philosophers do not know how to employ the ordinary epithets we use for describing 
mental activity. Philosophers are thus like the person who does not know how to 
employ the concept of a university. It is this para-mechanical model that Ryle attacks 
in his book. Its existence indicates that these theorists do not know how to wield the 
set of concepts that characterize our mental functions. 

The alternative he offers to the Official Doctrine is a detailed description of how 
mental concepts are used in everyday life. As he says: "The philosophical arguments 



which constitute this book are intended not to increase what we know about minds but 
to rectify the logical geography of the knowledge which we already possess." Ryle is 
thus reminding us of what we have always known, and also reminding us how philo- 
sophical conceits can blind us to the familiar. His description of the "logical geography" 
of mental concepts is thus a reminder of how we employ these concepts when we are 
not doing philosophy. Since any such employment is enormously complex, its "logical 
geography" will be lengthy, detailed, and specific. Here, by way of illustration, is a 
segment of a much longer specimen of logical geography: 

It is true that the cobbler cannot witness the tweaks that I feel when the shoe pinches. But 
it is false that I witness them. The reason why my tweaks cannot be witnessed by him is 
not that some Iron Curtain prevents them from being witnessed by anyone save myself, 
but that they are not the sorts of things of which it makes sense to say that they are wit- 
nessed or unwitnessed at all, even by me. I feel or have the tweaks, but I do not discover or 
peer at them; they are not things that I find out about by watching them, listening to them, 
or savouring them. In the sense in which a person may be said to have had a robin under 
observation, it would be nonsense to say that he has had a twinge under observation. There 
may be one or several witnesses of a road-accident; there cannot be several witnesses, or 
even one witness, of a qualm. (1949: 205) 

This passage is a good example of Ryle's way of exorcizing the ghost in the machine. 
The Official Doctrine presupposes that one has privileged access to a private realm 
consisting of one's own sensations, thoughts, and mental states; and that such an 
access consists in the observation of one's sensations and states. But to say that 
one is observing something implies that one is using one's eyes, or certain kinds of 
observational aids such telescopes, stethoscopes, and torches. One's eyes, and these 
instruments, can be used for the observation of planets, heart-beats, and moths. But 
we do not know what it would be like to apply them to felt sensations or to assert 
seriously that we "observe our pains." Since the Official Doctrine presupposes there is 
such a para-mechanical analogue as observing, it can be shown to be a species of non- 
sense by comparing its requirements with our actual use of such mental concepts as 
"tweaks" and "qualms." What the comparison reveals is a category mistake. The 
concept of observation applies to the physical domain in a way it logically cannot 
apply to the mental. Just as one logically cannot reach London gradually, so one cannot 
"observe" one's aches and pains. Ryle's line of reasoning throughout the work is thus 
to show that theorists have incorrectly wielded the ordinary concepts that describe 
human mental life. 

The Concept of Mind created a sensation when it appeared in 1949. For at least a 
decade after its publication it was the single most discussed book in Anglo-American 
philosophy. Nearly every periodical carried long articles about it. It was translated into 
a host of foreign languages, was taught in virtually every major western university, and 
within a short time seemingly had achieved the status of a philosophical classic. Yet a 
decade later it had fallen into obscurity, and subsequently it has hardly been referred 
to at all. What happened to occasion such a collapse? It is especially puzzling given that 
the book was of superb philosophical quality, was elegantly written, introduced many 
original and powerful distinctions, and was the first study to show in detail how the 



philosophy of language and the philosophy of mind are tied together. In this last 
respect, it was a bellwether for work that was to be developed thirty years later. 

There are several possibilities to explain what happened. One factor is that four years 
later Wittgenstein's posthumous Philosophical Investigations appeared (see WITTGEN- 
STEIN). It covered much the same territory as Ryle's study and in greater depth. As bril- 
liant as Ryle's book was it paled in comparison to the power and insight of 
Wittgenstein's. So philosophers turned from Ryle to Wittgenstein. It was the latter and 
not the former who was now read: Ryle had simply gone out of fashion. 

There is a second factor. Ryle claimed that in this work he was "charting the logical 
geography" of the many concepts used in speaking about the human mind. And 
though this was clearly an apt description it was also patent that his work had a strong 
verificationist thrust. Ryle frequently and in crucial passages speaks about the testabil- 
ity of propositions about mental concepts. For example, he states: "For, roughly the 
mind is not the topic of sets of untestable categorical propositions, but the topic of sets 
of testable hypothetical and semi-hypothetical propositions" (p. 46). Some critics have 
thus emphasized that Ryle's aim is to correct what other philosophers have said about 
the methods of verifying statements involving mental concepts, rather than trying to 
explicate these concepts themselves. The positivists, of course, identified the meaning 
of a statement with the method of its verification, and in many places in the Concept of 
Mind Ryle seems to presuppose that in describing how certain propositions involving 
mental concepts are to be tested he is explicating the meaning of those concepts. The 
book was thus eventually assessed as a sophisticated form of logical positivism, a view 
which had lost its influence by the 1950s. Ryle's work was swept away with the rest of 
this movement. 

Its behaviorism was a third factor. Ryle states that to give reasons for accepting or 
rejecting statements containing mental concepts will always involve hypothetical state- 
ments about overt behavior. In responding to the question, "What knowledge can one 
person get of the workings of another mind?" Ryle answers that it is "how we estab- 
lish, and how we apply, certain sorts of law-like propositions about the overt and the 
silent behavior of persons. I come to appreciate the skill and tactics of a chess player by 
watching him and others playing chess" (p. 169). Although Ryle always denied that he 
was reducing mind to behavior, and asserted instead that charting the "logical geog- 
raphy" of mental concepts was a philosophically neutral endeavor, his detailed analy- 
ses seemed to many philosophers to leave out one fundamental characteristic of the 
mind, the inward, felt quality of mental experience. For these philosophers such mental 
activities as deliberating or conjecturing, or such states as being in pain, were distinct 
from behavior. One could, for example, be in pain without evincing it in any mode of 
behavior. And even if one were to evince it, the pain itself was not to be identified with 
the behavior in question. A pain is not a grimace. So even if Ryle were correct in arguing 
that mental activity was exercised in various inter subjective situations it did not follow 
that the behavior so exhibited was identical with the mental events in question. Unlike 
Ryle, who minimized internal experience, Wittgenstein emphasized and acknowledged 
the existence of such phenomena. His point was that one should not identify them with 
such features as meaning, expecting, thinking, and so forth. And this position was seen 
to be more compelling than Ryle's. In the end this may have been the decisive factor in 
the eclipse of Ryle's reputation. 




Works by Ryle 

1931-2: "Systematically Misleading Expressions," Proceedings of the Aristotelian Society 32, pp. 

1945: "Philosophical Arguments," Inaugural Lecture as Waynflete Professor of Metaphysical 

Philosophy, Oxford. (Reprinted in Ryle 19 71, vol. 2, pp. 194-211.) 
1945-6: "Knowing How and Knowing That," Proceedings of the Aristotelian Society 46, 

pp. 1-16. 
1949: The Concept of Mind, London: Hutchinson. 
1950-1: "Heterologicality," Analysis 11, pp. 61-9. 

1954: Dilemmas, The Tamer Lectures, Cambridge: Cambridge University Press. 
1961: "Use, Usage, and Meaning," Proceedings of the Aristotelian Society, supp. vol. 3 5, 

pp. 223-30. 
1966: Plato's Progress, Cambridge: Cambridge University Press. 
1971: Collected Papers, vols. 1 and 2, London: Hutchinson. 

Works by other authors 

Lyons, W. E. (1980) Gilbert Ryle: An Introduction to his Philosophy, Brighton: Harvester. 
Wood, 0. P. and Pitcher, G. (eds.) (19 70) Ryle, Garden City, NY: Anchor Books. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

Alfred Tarski (1901-1983), Alonzo Church 
(1903-1995), and Kurt Godel (1906-1978) 


Alfred Tarski 

Tarski, born in Poland, received his doctorate at the University of Warsaw under 
Stanislaw Lesniewski. In 1942, he was given a position with the Department of 
Mathematics at the University of California at Berkeley, where he taught until 1968. 

Undoubtedly Tarski's most important philosophical contribution is his famous 
"semantical" definition of truth. Traditional attempts to define truth did not use this 
terminology and it is not easy to give a precise characterization of the idea. The under- 
lying conception is that semantics concerns meaning as a relation between a linguistic 
expression and what it expresses, represents, or stands for. Thus "denotes," "desig- 
nates," "names," and "refers to" are semantical terms, as is "expresses." The term "sat- 
isfies" is less familiar but also plausibly belongs in this category. For example, the 
number 2 is said to satisfy the equation "x 2 = 4," and by analogy we might say that 
Aristotle satisfies (or satisfied) the formula "x is a student of Plato." 

It is not quite obvious that there is a meaning of "true" which makes it a semanti- 
cal term. If we think of truth as a property of sentences, as distinguished from the more 
traditional conception of it as a property of beliefs or propositions, it turns out to be 
closely related to satisfaction. In fact, Tarski found that he could define truth in this 
sense in terms of satisfaction. 

The goal which Tarski set himself (Tarski 1944, Woodger 1956) was to find a "mate- 
rially adequate" and formally correct definition of the concept of truth as it applies to 
sentences. To be materially adequate a definition must "catch hold of the actual 
meaning of an old notion," rather than merely "specifying] the meaning of a familiar 
word used to denote a novel notion" (Woodger 1956: 341). Again, in discussing the 
material adequacy of some of his other definitions, Tarski writes, "Now the question 
arises of whether the definitions just constructed (the formal rigour of which raises no 
objection) are also adequate materially; in other words do they in fact grasp the current 
meaning of the notion as it is known intuitively?" (Woodger 1956: 128-9). 

To determine whether or not a proposed definition of a certain concept is materially 
adequate, Tarksi thinks that we must first formulate a criterion of material adequacy 
for such a definition: a precise condition which the definition must meet and which will 
guarantee that the defined notion is faithful to the original intuitive conception. Of 



course, whether a proposed condition really guarantees sufficient conformity to the old 
notion is subject to critical review. 

The requirement of formal correctness means that the proposed definition must be 
non-circular and that it must meet other logical constraints on acceptable definitions. 
One of the traditional requirements is that a definition must not define something in 
terms of things which are less clear than it. Tarksi even maintains that it must be speci- 
fied which previously adopted terms are to be used in giving the definition and requires 
that the formal structure of the language in which the definition is to be given be pre- 
cisely described. 

These are rigorous constraints. The motivating idea seems to be that only under such 
conditions can we hope to prove the material adequacy and formal correctness of a 
definition of truth. 

Tarski proposes as a criterion of material adequacy for a definition of truth that the 
definition shall have as logical consequences all instances of Schema (T): 

(T) X is true if and only if p, 

where "X" is replaced by a name of an arbitrary sentence of the language in question 
and "p" is replaced by that very sentence (or by a sentence with exactly the same 
meaning). The name in question must be a quotation-mark name or at least a name 
which necessarily designates the sentence. An appropriate instance of Schema (T) is 
thus such a thing as: 

(S) "Snow is white" is true if and only if snow is white. 

On the left-hand side of this "if and only if" there occurs a name of a certain sentence 
- which name is constructed by enclosing the sentence in question in quotation marks. 
Then using that name to mention the sentence, the property of being true is predicated 
of the sentence. On the right-hand side of the equivalence, the very sentence, which is 
named on the left and said there to be true, is used. The thing may appear to be a trivi- 
ality and perhaps that is all to the good. The condition, after all, is supposed to constrain 
an adequate definition in such a way that satisfying this condition guarantees that the 
definition catches hold of the actual meaning of the term "true." 

Note carefully that Schema (T) is not Tarski's definition of truth. That a definition 
should imply all instances of Schema (T) is the criterion of adequacy for the definition. 
But Tarski does seem to think that all the instances of (T) together completely capture 
the meaning of "true." If we could form an infinite conjunction, connecting all the 
instances with "and," we would have a complete specification of the semantical con- 
ception of truth. This is not an acceptable procedure according to the usual rules of 
definition, but a correct definition would be obtained if we could somehow achieve the 
same effect. 

Now the conditions which have already been given for an acceptable definition of 
truth require that the language involved be specified quite precisely. Natural languages 
do not have, or at least we do not know, rules which determine exactly what its expres- 
sions are; for example, the sentences of English are not precisely specified, ff we ignore 
this and set as our task to give a definition of truth for a natural language, say English, 
we encounter a paradox. No predicate of a sufficiently expressive language such as 
English can have the property that it validates every instance of Schema (T). And this 



is so whether the predicate is defined or not. The proof of this appeals to the infamous 
liar antinomy (or paradox). In a very simple version the antinomy (something "con- 
trary to law") goes like this. Consider 

(A) A is not true. 

That is, consider the sentence "A is not true," which sentence we have decided to name 
"A." Now Schema (T) implies: 

(1) 'A is not true" is true if and only if A is not true. 

But observing that the sentence A is the very sentence 'A is not true", we may assert: 

(2) A = 'A is not true." 

If two things are identical, then they share all the same properties. So, substituting the 
left-hand side of (2) for the right-hand side in (1), we get: 

(3) A is true if and only if A is not true. 

In the propositional calculus, this has the form: 

(4) P = ~P, 

"P if and only if not-P" and this is equivalent to the explicit contradiction: 

(5) P&-P, 

"P and not-P." 

Something must give. If we are unwilling to give up the usual laws of logic, since (2) 
is undeniable, it appears that we must alter or modify Schema (T), our criterion 
allegedly determined by the very meaning of "true." 

Tarski concludes, somewhat hastily, that ordinary language is inconsistent. The 
concept of truth must conform to Schema (T), but if we have such sentences as A, we 
arrive at a contradiction. The problem, says Tarski, is that natural languages are seman- 
tically closed, that is, they contain within themselves the terms and machinery for doing 
their own semantics. For example "is true in English" is itself a predicate of English. We 
must, he says, give our definition of truth in a metalanguage for the language whose sen- 
tences are in question. A metalanguage is a language which we may use to talk about 
another language. For example, in a book written in English which explains the 
grammar and meaning of the German language, the metalanguage is English. The lan- 
guage being studied is called the object language: in the case of this example, German. 
Further, claims Tarski, we must confine our attention to formalized languages which, 
unlike natural language, need not be semantically closed and which are otherwise 
precisely specified. 

With these provisos, Tarski proceeds to show that definitions of truth can be 
given for object languages which do not contain semantical terms. His method of defi- 
nition has the striking quality that the definition, given in a metalanguage, does not itself 
use any semantical terms. Because of the liar antinomy and other conundrums 
involving semantical notions, Tarski considered it important to give the definition in 
such a way that no semantical terms are presupposed as primitive or understood 
without definition. 



To see how the definition would be given for a very simple formalized language, let 
us suppose that we have just two predicates: "R," meaning is red, and "S," meaning is 
square. In addition, suppose that the language contains a variable x; a sign for nega- 
tion, say "-"; for conjunction, "&"; and a notation for universal quantification, "V" 
meaning For every. Thus, for example, we can write "Vx - S(x)" for "For every object x, 
x is not square" or, more naturally, "Nothing is square." 

We assume that our metalanguage contains the means of expressing at least the 
very same notions as the object language. Here we are using a bit of English as meta- 
language so that we have the words "is red" to mean the same as the predicate "R" in 
the simple formalized language. Now let some domain of objects be selected as the col- 
lection of things we will be talking about. One can then define satisfaction for the object 

(1) An object satisfies "R(x)" if and only if it is red. 

(2) An object satisfies "S(x)" if and only if it is square. 

(3) An object satisfies a negation >—$ if and only if it does not satisfy <|>. 

(4) An object satisfies a conjunction of the form ty & \|/l if and only if it satisfies § 
and it satisfies \|/. 

(5) An object satisfies a universal quantification ^V x ty 1 if and only if every object 
(in the domain) satisfies <|>. 

Here <|> and \|/ are formulae of the formalized language. These are expressions which we 
have not really defined but which include such things as "R(x)" ("x is red"), "— [S(x) & 
R(x)]" ("It is not the case that x is square and x is red"), as well as sentences such as 
"— Vx - S(x)" ("It is not the case that for every x, x is non-square," i.e. "Something is 

This doesn't look like a definition, but in fact it really does completely explain the 
meaning of "satisfies" as it applies to our simple language. Using these definitional rules 
on complicated expressions we can proceed step by step to simpler expressions until we 
get down to cases covered by (1) and (2). And it may look as if we have some kind of 
vicious circularity. For example, we have used "and" (in the metalanguage) to define 
satisfaction for expressions (of the object language) containing "&." But the appear- 
ance is deceptive. We have assumed that whatever we can say in the object language, 
we can say in the metalanguage, but not necessarily vice versa. This assumption does 
not introduce any logical or philosophical difficulty into the definition. 

Finally we define truth: 

A sentence § is true if and only if every object (in the domain) satisfies it. Again, we 
haven't really defined the sentences of our object language, but they will be expressions 
in which no occurrences of the variable are "dangling." For example, "— Vx — [R(x) & 
S(x)]" ("Something is red and square") is a sentence, as opposed to a formula such as 
"R(x)" ("x is red"). Here the variable x is just a placeholder, indefinitely indicating some- 
thing or other, but no definite thing. 

It is not obvious that this definition actually conforms to the criterion of material 
adequacy. But it does. It can be proved that every instance of Schema (T), confined to 
sentences of our object language, is a consequence of this definition. The whole thing 
may seem trivial, but it is really quite amazing that in an appropriate metalanguage 
truth can be defined without appealing to any semantical notions. This means that it 



has been defined in terms of things which are clearer: they are just the concepts of logic 
together with the concepts of the object language. 

It remains to mention Tarski's work on the notion of logical consequence (Woodger 
1956: 409-20). This, like the notion of truth, was used in an intuitive way by logicians 
and philosophers before Tarski, but it was the latter who made the notion precise. 

Consider once again our simple formalized language. Do not select a particular 
domain and particular meanings for "S" and "R." Rather, contemplate any arbitrary 
interpretation of the language - any domain of objects whatsoever and any appropriate 
meanings for these symbols. The logical symbols "-," "&," and so on, are to retain their 
original meanings throughout. 

For any such specification, we can explain truth under that interpretation along the 
lines used above for the particular interpretation we were considering. Suppose that in 
some interpretation a particular sentence, say "Vx[S(x) & R(x)]," comes out true. Then 
in that interpretation certain other sentences will come out true as well. For example, 
"VxS(x)" and "VxR(x)." In fact, this will always happen. If an interpretation makes our 
example sentence true, that interpretation will also make these two sentences true. In 
such a case, Tarski says that the latter two sentences are logical consequences of the first 
sentence. In general, a sentence i|/ is a logical consequence of a sentence <|> if and only 
if every interpretation which makes <|> come out true also makes v|/ come out true. And 
a sentence is defined to be a logical consequence of a collection, or set, of sentences if 
every interpretation which makes every sentence in the set come out true also makes 
the sentence in question come out true. Finally, Tarski defines a sentence to be logically 
valid if it comes out true under every interpretation. 

The importance of such a definition is that we can now strictly define what it is for 
something to be a valid argument in our language. And, of course, the study of valid 
arguments is at the very heart of the discipline of logic. Using these definitions we can 
then prove that certain systems of logical rules are "complete" in the sense of being 
adequate to their intended purpose of capturing all valid inferences expressible in the 
language. For example, certain formulations of first-order logic, the logic of such 
notions as and, not, or, if, . . . then, not, some all, and the like were proved complete by 
Kurt Godel, to be discussed below. 

These two things, his definition of the concept of truth for formalized languages and 
his explication of the concept of logical consequence are Tarski's distinctive philo- 
sophical contributions. They are substantial indeed. 

Alonzo Church 

In 1927, Church received his Ph.D. from Princeton, where he taught from 1929 to 
1967. Thereafter, he taught at UCLA until 1990. He was a long-time editor of the 
Journal of Symbolic Logic, which he helped to found. Church's philosophical contribu- 
tions largely concern questions about the foundations of logic and mathematics, espe- 
cially their ontology, and topics in the philosophy of language and in the related area 
of intensional logic. 

Church's thesis is a hypothesis concerning the identification of the mechanically 
computable or calculable functions discussed below in connection with Godel's incom- 
pleteness theorem. Church proposed as a precise mathematical analysis of the idea of 



such functions that they be identified with the lamb da- definable functions. This latter 
notion is too technical to be explained in detail here. Alan Turing independently pro- 
posed the identification of the mechanically computable functions with functions com- 
putable in principle by a precisely definable sort of abstract "machine," now called a 
Turing machine. This identification turned out to be equivalent to Church's thesis. That 
is, the class of lambda-definable functions is exactly the same as the class of functions 
computable by a Turing machine. Other attempts to analyze the notion in question have 
always led to the same class of functions. The identification of the class of mechani- 
cally ("algorithmic ally") calculable functions with the class of lambda-definable or 
Turing machine-computable functions (the "Church-Turing thesis") is now almost 
universally accepted. 

Church's theorem, to be carefully distinguished from Church's thesis, is a theorem of 
mathematical logic to the effect that there is no effective (= mechanical) procedure for 
deciding whether or not a formula of first-order logic is valid. 

Church was a Platonist or, as he preferred, a realist about the entities apparently 
described and studied by mathematics and logic. Numbers and other mathematical 
entities are, he believed, objectively existing, mind-independent objects and mathemat- 
ics itself consists of truths about these things. Logic seems to require, if formulated in 
full generality, propositions, properties, and "individual concepts." These kinds of things, 
usually called intensional entities, are supposed to be abstract, real, and objective enti- 
ties suitable to be the meanings of expressions in various languages. Propositions, for 
example, are claimed to be the meanings of declarative sentences, the same for syn- 
onymous sentences, whether in a single language or in two or more different languages. 

Church's general methodological viewpoint about the formal sciences was a kind of 
"hypothetico-deductive rationalism." According to this view, intuitions or feelings of 
self-evidence provide initial support for assumptions about abstract entities. The theo- 
ries of these are to be formalized, stated using the precise language and terminology of 
symbolic logic, and the results are to be evaluated using the sorts of criteria common 
to scientific procedures in general. One way we evaluate theories is by deducing conse- 
quences and thereby determining whether they are adequate to account for the data. 
In the formal sciences Church took the data to include the accepted facts of mathe- 
matics and logic. 

Many of Church's philosophical contributions appear in reviews in the Journal of 
Symbolic Logic. His relatively few papers devoted explicitly to philosophical topics 
usually concerned questions about meaning and related topics in the philosophy of lan- 
guage. There are also arguments against nominalism as it is sometimes espoused in 
connection with mathematics, logic, or semantics. 

As a sample of the latter (Church 1950), consider a nominalist attempt to give an 
analysis of certain statements apparently about propositions. Suppose it is claimed that 
such a sentence as (1) "Seneca said that man is a rational animal" is to be analyzed as: 
(2) "There is a language S' such that Seneca wrote as a sentence of S' words whose 
translation from S' into English is 'Man is a rational animal'." This may already seem 
excessively complicated, but simpler attempts to analyze statements about assertion so 
that they concern such relatively concrete things as sentences are subject to easy refu- 
tation. To bring out clearly that (2) will not do as an analysis of (1), Church uses the 
"translation test," a procedure whose invention is usually attributed to C. H. Langford. 



If we translate (1) into German, we get (1') "Seneca hat gesagt, das der Mensch ein 
verniinftiges Tier sei". In translating (2) into German, note carefully that the word 
"English" must be translated as "Englisch" (not as "deutsch") and the quotation which 
forms part of (2) is to be translated as "Man is a rational animal" (not as "Der Mensch 
ist ein verniinftiges Tier"). This latter translation, call it (2'), certainly would not convey 
anything like the information which would be conveyed to a German speaker (who 
spoke no English) by (1'). Thus, argues Church, (1') is not an acceptable analysis of (1). 
The basic idea of the objection, which can be seen even without appealing to transla- 
tion, is that (1) does not say anything about any particular language (and so neither 
does its translation (1')). whereas (2) makes specific reference to English. 

A philosophical argument which has a quite surprising conclusion is given by 
Church (Church 1956: 24-5) as a more precise version of reasoning offered by Gottlob 
Frege. The conclusion of the argument is that sentences denote truth-values, true sen- 
tences denoting Truth (or The True) and false sentences denoting Falsehood (or The 
False)! Put like this, the thesis seems quite incredible, even unintelligible. Why suppose 
that sentences "denote" anything at all? And what, we may ask, are these alleged 
"objects," Truth and Falsehood? These are good questions, but the essential point of 
Church's argument (and Frege's before him) could be stated like this: the truth or falsity 
of a sentence is the only thing that stands to the sentence as the denotation of a 
(complex) name stands to its parts. To see this take such a sentence as (a) "Sir Walter 
Scott is the author of Waverley." If we replace "the author of Waverley" by an expres- 
sion which denotes the same, "the man who wrote twenty-nine Waverley Novels alto- 
gether," we get a new sentence: (b) "Sir Walter Scott is the man who wrote twenty-nine 
Waverley novels altogether." If we are supposing that the "denotation" of a sentence, 
whatever it is, is unchanged if a denoting part is replaced by another with the same 
denotation, then this new sentence must have the same denotation as the original. 
Further, it is plausible (Church claims) that the sentence, (c) "The number, such that 
Sir Walter Scott wrote that many Waverley Novels altogether is twenty-nine," is so close 
in meaning to (b) as to have the same "denotation" (again, without yet assuming that 
we know what this is). But now let us replace the denoting expression "The number, 
such that Sir Walter Scott wrote that many Waverley Novels altogether" in (c) by an 
expression with the same denotation, namely; "The number of counties in Utah" 
(which is in fact twenty-nine). We then get a sentence which is supposed to have the 
same denotation as (c), (d) "The number of counties in Utah is twenty-nine" (again 
assuming that a sentence does not change its denotation if a denoting part is replaced 
by another with the same denotation). 

Now compare our original sentence (a) "Sir Walter Scott is the author of Waverley" 
with (d) "The number of counties in Utah is twenty-nine." By the reasoning just 
explained, these two sentences must have the same "denotation." But the only 
meaning-relevant feature which they seem to have in common is that both are true. A 
little reflection on such examples points to the conclusion that the only thing that can 
be expected to remain invariant under such substitutions is the truth or falsity of the 
original sentence. So if "denotation" has an analog for sentences, it will have to be the 
truth-values, truth and falsity, which may be seen as mathematical abstractions. 
(Compare the mathematical abstraction of numbers, as objects, from collections or 
from concepts of collections.) The Church-Frege argument here may not be conclusive, 



but the analogy uncovered is striking and it may well be a useful theoretical assump- 
tion for semantics that sentences "denote" truth-values. (See Anderson 1998 for 
further discussion.) 

Church's most important philosophical ideas are contained in his work on the foun- 
dations of intensional logic (Church 1951, 1973, 1974). Philosophers and logicians 
contrast intension and extension, but it is by no means easy to give a clear characteri- 
zation of these notions. In the case of sentences, Church would maintain that the sense, 
or intension, of the sentence is the proposition which it expresses and the denotation, 
as already explained, is the truth-value of the sentence. Logic as standardly taught in 
philosophy and mathematics departments makes no significant distinction between 
sentences with the same truth-value; arguments which turn on finer distinctions of 
meaning are simply not treated. Similar distinctions hold between the set of things of 
which a predicate is true, the extension of the predicate, and the property conveyed by 
the predicate, its intension. Again, a distinction between the meaning, strictly so-called, 
of an expression such as "The present president of the US" (its intension) and what it 
stands for, the actual person, is needed. Here we might say again with Church, that the 
meaning of the expression in the strict sense is the concept that it expresses, its inten- 
sion, but what it denotes or stands for, the person or, more generally, the object, is its 

So, as already explained, Church calls the proposition expressed by a sentence its 
sense and the truth-value that it stands for its denotation. Predicates have properties as 
their senses and sets as their denotations, and individual expressions (e.g. descriptive 
names) have certain concepts as their senses and what they stand for as their denota- 
tions. The relationship that holds between the sense of an expression and what it 
denotes let us call the concept relation, and symbolize it by the capital Greek letter A 
(delta). Then propositions are concepts of truth-values, properties are concepts of sets, 
and individual concepts are concepts of the individual things that the concepts char- 
acterize. Generalizing our terminology (as Church does), call anything that is capable 
of being the sense of some expression a concept. 

The intensional logic that Church envisioned would have two kinds of intensional 
axioms: logical principles about A and principles that would specify the essential char- 
acteristics of propositions and other complex concepts. In connection with the latter, 
Church took it to be especially important to have axioms which give, or correspond to, 
criteria of identity for complex concepts. A criterion of identity in the present case is a 
principle that determines the identity or difference of the complex concepts expressed 
by different sentences (or predicates or descriptive names) in terms of some known rela- 
tion between the sentences (complex expressions) themselves. An example would be the 
principle that two sentences express the same proposition if and only if they are logi- 
cally equivalent; in our example, that is, they have the same truth-value necessarily, or 
on logical grounds alone. 

We have already explained that a function of numbers is a correlation of a certain 
kind. Thus, square or squaring is said to be a function from numbers to numbers. In 
general, any correlation between the things in two collections is called a function. 
Generally, a function is just any conceivable correlation between the things in one col- 
lection and the things in another (or, possibly, the same) collection; it is allowed that two 
or more things in the first collection be correlated with the same thing in the second. 



A name of a function has both a sense and (in general) a denotation. The sense 
is therefore a concept of the function denoted by that name. For example the 
expression "The squaring function" denotes the function that takes each number into 
its square and it has as its sense what is conveyed by "x 2 ." The combination of a name 
of a function with the name of something to which the function is applied (an 
argument of the function) will also have a sense: a complex sense involving the sense of 
the function name and the sense of the name of the entity to which the function is 
being applied. 

The importance of this idea appears in the observation that any complex expression 
may be construed as being built up from a function expression, together with expres- 
sions for one or more arguments to which the function is applied. 

Now let us write "A(X, Y)" to mean that X is a concept of Y The axioms which Church 
took to govern the delta-relation are: 

(CI) For every X, Y, and Z, if A(X,Y) and A(X,Z), then Y = Z. 

(C2) For every F and F lt if A(F lt F), then for every X and X lf if A(Xj,X), then 

A(F 1 X 1 ,FX). 
(C3) For every F and F lt if for every X and X lf A(X,X 2 ) implies that A(F 1 X 1 ,FX), then 

A(F 1) F). 

(CI) says that anything which is a concept of something is a concept of exactly one 
thing. In (C2) and (C3), F is any function and FX is the result of applying that function 
to an argument X; that is, FX is the entity which is correlated with X by the function. 
Where F : and X : are concepts, we have just written "FjXj" for the complex concept that 
results when the concept F : is combined with the concept X 1 . In these terms, (C2) 
amounts to the claim that if an expression denoting a function is combined with an 
expression denoting an argument (in some possible language), then the sense of the 
complex expression is the result of combining the sense of the function name with the 
sense of the argument name. 

The proposed axiom (C3) is more problematic. To understand and accept it, one 
really must go along with a hypothesis that Church proposes to simplify the logic of the 
system. Church assumes that a concept of a function can be taken to be a function from con- 
cepts to concepts. This is fine for axiom (C2), which is then just understood in such a way 
that combining a concept of a function with a concept of an argument is nothing more 
than applying a certain kind of function to a certain kind of argument. But axiom (C3) 
is much bolder. It amounts to the claim that any function from concepts to concepts 
satisfying a certain condition is a concept of a certain function. It says: if a function 
applied to a concept of an argument always yields a concept of the output of some func- 
tion applied to the argument thus concepted, then the function (from concepts to con- 
cepts) is a concept of the function from objects to objects. 

This axiom leads to various difficulties, which cannot be explained here (see 
Anderson 1998). It is fair to say that even the basic principles of intensional logic, as 
Church conceived it, are still not settled. 

Intensional principles of the second sort - those supposed to individuate 
complex concepts - are also still problematic. Church proposed three heuristic 
principles to guide the formulation of such axioms: (A) that logically equivalent 
expressions express the same concept, (B) that expressions that have exactly the same 



syntactical structure and whose corresponding parts have the same meanings express 
the same concept, and (C) expressions that can be obtained from one another by apply- 
ing the logical operation of lambda conversion express the same concept. Lambda con- 
version is a logically valid transformation of expressions, which we do not attempt to 
explain here. 

The idea behind (A) works well if, but only if, one is dealing with reasoning involv- 
ing no finer distinctions of meaning than are involved in arguments turning on modal- 
ity: necessity, possibility, impossibility, and similar conceptions. Suggestion (C) appeals 
to the technical notion of lambda-conversion which is very difficult to motivate from a 
philosophical point of view. Hands down, the notion urged in (B) is the most promis- 
ing. Church (who agreed with the assessment just offered) tried to implement this 
approach several times in his published work, but technical and logical difficulties still 
block the way of a satisfactory theory. 

ft is fair to say that this project to which Church contributed fundamental and 
important work, to establish a comprehensive and adequate general intensional 
logic, has not yet been completed. But his successful philosophical contributions are 
impressive indeed. 

Kurt Godel 

Kurt Godel received his doctorate in 1930 at the University of Vienna. He emigrated to 
the United States in 1940 and soon afterwards became a member of the Institute for 
Advanced Studies at Princeton, New Jersey, until his death. Godel, like Tarski and 
Church, is best known as a logician. But his logical discoveries are of profound signifi- 
cance for parts of philosophy: the philosophy of logic and mathematics, epistemology 
and (perhaps) the philosophy of mind. In addition, in later years Godel concentrated 
on philosophical questions and made strikingly original suggestions as to their solu- 
tion, including an improved version of Anselm's famous ontological argument for the 
existence of God, as elaborated by Leibniz. 

Godel's most famous discoveries are his two incompleteness theorems (Godel 1931). 
Here we will give outlines of modernized version of his proofs. Suppose that the arith- 
metic of the natural numbers (0, 1, 2, and so on) is formulated as an axiomatic system. 
The language used is a precisely specified symbolic, or formalized, language with 
axioms stating the basic properties of the natural numbers and rules of inference 
stating which sentences may be correctly inferred from others. A sequence of sentences 
beginning with axioms and constructed by applying the rules of inference is said to be 
a proof of the last sentence in the sequence, which latter is said to be a theorem of the 

Godel observed that we can assign numbers (now called "godel numbers") to 
the syntactical entities of the axiomatic system. That is, one can correlate numbers 
with symbols, with complex expressions, and even with sequences of expressions such 
as proofs. These numbers are assigned to symbols of the object language in the 
metalanguage. But these numbers are part of the subject matter of the object language 
theory itself. 

This done, we have a sort of indirect way of talking about expressions and sequences 
of expressions of the formal theory within the theory itself. By talking about the godel 



numbers of expressions and sequences of expressions, we can simulate, or model, talk 
about the expressions and sequences. 

Godel then proved that if the formal system of arithmetic meets certain minimal con- 
ditions of adequacy, then the set of godel numbers of the sentences provable in the 
system (the theorems of the system) can be defined within that system. The condition 
of minimal adequacy is that the system of arithmetic be capable of expressing certain 
functions of natural numbers. A function of natural numbers is just a correlation 
between numbers and numbers, or between pairs of numbers and numbers, or . . . and 
so on. Intuitively, the functions of natural numbers that must be expressible are those 
whose values for given arguments can be "effectively calculated": calculated by means 
of an algorithm or recipe, mechanically (and mindlessly) computed in a manner avail- 
able to a computing machine. 

Next, it can be proved that the set of godel numbers of true sentences ("true" being 
defined in the manner of Tarski) is not definable in arithmetic. The proof uses an argu- 
ment which parallels the reasoning of the liar antinomy (!) but which is logically unex- 
ceptionable. The conclusion is that such a system of arithmetic cannot contain or define 
its own truth predicate (as applied to godel numbers as surrogates for sentences). 

But if the set of godel numbers of provable sentences is definable in arithmetic and 
the set of godel numbers of true sentences (of arithmetic) is not definable in that system, 
then the two sets have to be different. Therefore, either some sentence provable in arith- 
metic is not true or some true sentence of arithmetic is not provable. We cannot accept 
the former, at least if we have chosen a system of axioms which we can see to be true 
of the natural numbers. We conclude that some true sentence of arithmetic is not a 
theorem of arithmetic. Any formal system of arithmetic meeting reasonable conditions 
of adequacy will be incomplete. This is essentially Godel's first incompleteness theorem. 

Godel's actual proof did not proceed in the way we have described. Rather, he 
assumed not that the axioms of arithmetic are true (and that the rules of inference pre- 
serve truth) but only that arithmetic is consistent: it is not possible to derive an actual 
contradiction from the axioms using the rules of logic. (Really, he assumed that arith- 
metic is omega-consistent, a stronger assumption than consistency which we need not 
explain here.) Then Godel showed how to construct, given that arithmetic is consistent, 
a particular sentence G which is such that neither G nor its negation -G ("not-G") is a 
theorem. The proof proceeds in such a way that we could, with sufficient patience and 
longevity actually write down a true sentence of arithmetic which, if arithmetic is 
formally consistent, cannot be proved in arithmetic. And one can see that if the con- 
sistency of arithmetic is accepted, the sentence G is the true but unprovable one, as 
opposed to -G. 

This sentence G involves just quite ordinary arithmetical concepts such as "plus" and 
"times" together with the usual logical concepts "and," "not," "some," "all," "equals," 
and so on. It is worth noticing that, contrary to various popular expositions, Godel's 
original proof does not involve self-reference in any sense. The true but unprovable sen- 
tence G does not "say" that it, itself, is unprovable. It is a sentence entirely about natural 
numbers and their properties and relations. But it is a sentence that simulates such a 
self-referential sentence in the sense that it is true if and only if it is not provable. 

Well, so what? It is natural to suggest that the axioms of arithmetic with which we 
began are just not adequate and that some new axioms must be added. It is as if Euclid's 



geometry had been formulated without the Parallel Postulate. One would simply have 
to add it, or some equivalent. However, if you consider the details of Godel's proof, it is 
evident that a system obtained by adding any finite number of axioms will still be 
subject to the proof - although the unprovable but true sentence will be different. 
Indeed, even adding infinitely many new axioms in any "effective" way does not evade 
the proof. We must conclude that nothing that would count as a formal system can 
contain all the truths of arithmetic. (If you think that to be a truth of arithmetic is to 
be provable in arithmetic, then this result will be quite difficult to comprehend. But one 
conclusion we can draw from these considerations is that this identification cannot be 

This much is already quite startling. The goal of mathematicians since Euclid 
has been to specify certain basic truths of mathematics and to justify all others 
by deduction from these. Godel proved that this goal is unattainable! No matter 
what formal (alias "axiomatic") system is proposed, there will be truths of arith- 
metic (the most basic part of mathematics) which the system cannot prove. Of 
course to show this with mathematical precision requires that we precisely 
define "axiomatic system" or "formal system," but this can be done in a way that is 
undeniably correct. 

Godel's second incompleteness theorem builds on the first. Using the technique of 
pseudo-self-reference mentioned above, one can find a sentence of any (minimally ad- 
equate) formal system of arithmetic which "says" that the system itself is consistent. 
Call this sentence "Consis." Now the proof of Godel's first incompleteness theorem can 
be mimicked within arithmetic to produce a proof of the conditional sentence: "If 
Consis, then G." Suppose it were possible to prove within arithmetic the sentence Consis 
which mirrors the proposition that arithmetic is consistent. Then it would be possible 
to prove (by modus ponens) the Godel sentence G. But we already know, from the first 
incompleteness theorem, that if arithmetic is consistent, G cannot be proved therein. If 
we suppose, as we are certainly entitled to do, that the theorems of arithmetic are true, 
then arithmetic is consistent. We conclude that the sentence which (in an indirect 
sense) expresses that arithmetic is consistent, cannot itself be proved in arithmetic. 
And, of course, the sentence "expressing" that arithmetic is inconsistent will not be a 
theorem either. It too, like the sentence G, is undecidable in arithmetic: can neither be 
proved nor refuted therein. 

We have been at some pains to dispel the impression that Godel's proofs 
literally involve self-reference. What then does, for example, the sentence "expressing" 
the consistency of arithmetic look like? Well, like the sentence G, it is written 
entirely in the language of arithmetic (involving "plus," "times," and "equals," for 
example). In fact it can be seen as expressing a certain mathematical claim about a 
polynomial. Let P(x,y) be a polynomial involving just the two indicated variables 
and integral coefficients. Then the statement "For every y, there is an x, such that 
P(x,y) = 0" may be true or false, depending on the details of the polynomial. Problems 
of this sort, as to whether or not such a statement is correct, are called "diophantine" 

problems. More generally, suppose we have n variables x lt x 2 x n and m variables 

y lt y 2 y m , and a polynomial (with integral coefficients) involving these, say 

"P(x! x n , y : y m )-" Then the statement "expressing" the consistency of arith- 
metic is of the form: 



For every y : y m , there are x : x n , such that P(x : x n , y : y m ) = 0. 

This is a purely arithmetical statement and, one would have supposed, a claim that 
arithmetical techniques should be able to settle, yea or nay. But they cannot. (The 
details of these proofs are rather difficult to grasp. The best introduction to them is 
Smullyan 1957. For precise details, Boolos and Jeffrey 1989 is excellent.) 

The philosophical import of these results is more controversial. We have already 
observed that a certain mathematical program is thereby proved impossible: that of 
deducing all of mathematics from an axiomatic basis. If we think of metaphysics as 
including all necessary truths, mathematics not excluded, then the goal, perhaps most 
closely associated with Spinoza, of an axiomatic development of metaphysics is thus 
also proved impossible to achieve. 

Godel's own philosophical speculations on the import of his theorems are most 
clearly articulated in his Collected Works (1995: 304-23). Let us confine our attention 
to arithmetic and speak of the true sentences thereof as "objective mathematics." 
Human beings, using our presently accepted arithmetical assumptions, can certainly 
prove some of these sentences. Godel calls the mathematical truths that human beings 
are capable of demonstrating "subjective mathematics" (perhaps not the best choice of 
terminology since these are all objectively true and indeed knowable to be such). Some 
of these may require axioms about the numbers which are presently unknown, but 
which can in principle be seen to be evident by human mathematicians. 

Now consider again Godel's second incompleteness theorem. The theorems of 
a formal system of arithmetic comprise a set of sentences that could be mechanically 
generated, one after the other. According to Godel's second incompleteness theorem, 
any such system will be unable to prove (generate) the arithmetical sentence that 
"expresses" its own consistency. Recall that this was a certain sentence expressing 
a diophantine arithmetical problem. Godel draws the following conclusion from 

Either mathematics is incompletable in this sense, that its evident axioms can never be 
comprised in a finite rule, that is to say the human mind (even within the realm of pure 
mathematics) infinitely surpasses the powers of any finite machine, or else there exist 
absolutely unsolvable diophantine problems of the type specified. (1995: 310) 

To say that the evident axioms (and rules of inference) can be "comprised in a finite 
rule" is equivalent to the possibility of the formulation of arithmetic as a whole as a 
formal system. And this latter amounts to the possibility of being generated in a 
machine-like fashion. To say that a diophantine problem is absolutely unsolvable means 
that neither the statement nor its negation will ever be a theorem of subjective math- 
ematics. There is no doubt that this conclusion actually does follow from Godel's proof, 
together with the mentioned analysis of a finite mechanical procedure. Godel himself 
thinks, and argues, that the second option is incorrect - there are no absolutely unsolv- 
able arithmetical problems - but his arguments for this conclusion are not airtight and 
may be reasonably doubted. 

Godel also had an improvement on Anselm's ontological argument for the existence 
of God, especially as it was developed by Leibniz. Leibniz observed that the one 



thing that needs to be proved to complete Anselm's proof is that the existence of God 
ispossible. A number of commentators on the argument, including Kant, have observed 
that all that is really established by Anselm is that if God exists, then He necessarily 
exists. Now, given the nature of the proof, we may further conclude that this condi- 
tional itself is necessary. So, we may conclude further, by a standard principle of modal 
logic, that if it is possible that God exists, then it is possible that it is necessary that He 
exists. (The standard principle of modal logic in question is this: ifp necessarily implies 
q, then if p is possible, then q is possible.) If we further suppose that we can somehow 
prove that it is possible that God exists (given a certain definition of "God"), then it 
follows that it is necessary that it is possible that God exists. But according to one plau- 
sible system of modal logic (standardly called "S5"), it then follows that it is necessary 
that God exists! 

But can we prove that it is possible that God exists? Godel thought that we can 
and that a version of the ontological argument is then cogent. His argument for 
this, which bears some resemblance to Leibniz's argument for the same conclusion, 
uses the idea of a positive property. Godel doesn't really say very clearly what this 
conception involves, but he remarks that it has two possible interpretations, "positive 
in the moral-aesthetic sense" and positive in the sense of involving only "pure 
attribution." A being is defined to be God-like if it has every positive property. Then a 
property is defined to be an essence of an entity x if x has that property and it entails 
every other property that x has. An entity necessarily exists, by definition, if every 
essence of it is necessarily exemplified. Finally, Godel assumes the following "axioms" 
about these concepts: 

1 A property is positive if and only if its negation is not positive. 

2 Any property entailed by a positive property is itself positive. 

3 The property of being God-like is positive. 

4 If a property is positive, it is necessarily positive. 

5 The property of necessarily existing is positive. 

From these it follows that it is possible that a God-like being exists and, by essentially 
the argument explained earlier, that such a being therefore exists, and indeed neces- 
sarily so. There are some problems with the argument, not the least of which is the 
obscurity of the notion of a positive property. For an able discussion of the argument 
and its alleged defects, see Adams 1970. 

Godel also thought that Einstein's Theory of Relativity has implications for idea- 
lism, in particular that it supports some ideas of Immanuel Kant. He argues that 
there is considerable reason to believe that "time is unreal" (1951: 555-62). Essentially 
the argument is this: If time and change are something real, then there must be 
such a thing as an objective and absolute lapse of time. But the Theory of Relativity, a 
well-confirmed scientific theory, seems to deny that there is such an objective 
lapse. Godel considers various objections and explains the relevance of a technical con- 
tribution of his to that theory. That work, by the way, seems to imply the possibility of 
"time travel"! 

In sum, we may say that Godel's main work in logic is of profound philosophical 
significance and that his other philosophical work certainly deserves further 
careful study. 




Works by Church, Godel, and Tarski 

Church, A., 1950: "On Carnap's Analysis of Statements of Assertion and Belief," Analysis 10, 

pp. 97-9. 
1951: "A Formulation of the Logic of Sense and Denotation," in Structure, Method and Meaning, 

ed. P. Henle, H. M. Kallen, and S. K. Langer, New York: Liberal Arts Press. 
1956: Introduction to Mathematical Logic, vol. T, Princeton, NJ: Princeton University Press. 
1973: "Outline of a Revised Formulation of the Logic of Sense and Denotation (Part I)," Nous 7, 

pp. 24-33. 
1974: "Outline of a Revised Formulation of the Logic of Sense and Denotation (Part II)," Noiis 

8, pp. 135-56. 
Godel, K., 1931: "Uber formal unentscheidbare Satze der Principia Mathematica und verwandter 

Systeme I," Monatshefte fur Mathematik undPhysik 38, pp. 173-98. 
1951: "A remark about the relationship between relativity theory and idealistic philosophy," in 

Albert Einstein: Philosopher-Scientist, ed. P. A. Schilpp, New York: Tudor Publishing Company, 

pp. 555-62. 
1986: Collected Works, vol. I, ed. S. Feferman, J. Dawson, and S. Kleene, New York: Oxford 

University Press. 
1995: Collected Works, vol. Ill, ed. S. Feferman and J. Dawson, New York: Oxford University Press. 
Tarski, A., 1944: "The Semantic Conception of Truth and the Foundations of Semantics," 

Philosophy and Phenomenological Research 4, pp. 341-76. 
1956a: "On Definable Sets of Real Numbers," in J. H. Woodger, Logic Semantics, Metamathematics: 

Papers from 1923 to 1938 by Alfred Tarski, Oxford: Clarendon Press, pp. 110-42. 
1956b: "On the Concept of Logical Consequence," in Woodger, pp. 409-20. 
1956c: "The Concept of Truth in Formalized Languages," in Woodger, pp. 152-2 78. 

Works by other authors 

Adams, R. M. (1995) "Introductory note to *1970," in Kurt Godel: Collected Works, vol. Ill, ed. 

S. Feferman, et al., Oxford: Oxford University Press, pp. 388-402. 
Anderson, C. A. (1998) "Alonzo Church's contributions to philosophy and intensional logic," 

Bulletin of Symbolic Logic 4, pp. 129-71. 
Boolos, G. S. and Jeffrey, R. C. (1989) Computability and Logic, 3rd edn., Cambridge: Cambridge 

University Press. 
Smullyan, R. (1957) "Languages in which self reference is possible," Journal of Symbolic Logic 

22, pp. 55-67. 
Woodger, J. H. (trans.) (1956) Logic Semantics, Metamathematics: Papers from 1923 to 1938 by 

Alfred Tarski, Oxford: Clarendon Press. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

F.P.Ramsey (1903-1930) 


Frank Plumpton Ramsey made lasting contributions to philosophy, logic, mathematics, 
and economics in an astonishingly short period. He nourished during the 1920s at 
Cambridge University, and he interacted with the many notable figures there, includ- 
ing Russell, Moore, Keynes, and Wittgenstein. He was by no means a minor figure 
among this group. His work makes it clear, in fact, that he was at least their intellec- 
tual equal, and perhaps more, a judgment quite consistent with opinions expressed 
by his contemporaries. 

Ramsey held an appointment in mathematics, but his main mathematical interests 
were in its foundations. He published just one piece of real mathematical work, a nine- 
page first section of his investigation of a decision procedure for a special case of first- 
order predicate logic ("On a Problem of Formal Logic," in Ramsey 1931). What he 
presented there as a preliminary tool has since been recognized as an important result 
- Ramsey's Theorem - and is the origin of a now thriving branch of mathematics, 
Ramsey Theory. He also published two papers in economics, one on taxation, and 
another on saving, that have also come to be regarded as important pioneering contri- 
butions to their subjects (in Ramsey 1978). 

Ramsey had wide philosophical interests. He criticized and revised the logical system 
of Principia Mathematica, simplified its theory of types, and distinguished between the 
logical and semantic paradoxes. He gave a proto-functionalist account of belief, 
together with a redundancy theory of truth. He developed now influential accounts of 
partial belief, reasonable belief, probability, and knowledge. He developed an account of 
causality closely related to Hume's that has strongly influenced important contempo- 
rary accounts of laws of nature, and he made a proposal for representing the content 
of scientific theories via what are now known as Ramsey sentences. This is a distin- 
guished collection of original contributions to philosophy; it is astonishing in light of 
the tragic fact that Ramsey died shortly before his twenty-seventh birthday, in January 
1930. The circumstances of Ramsey's short life, and the time it took for many of his 
various projects to influence others, put us in an unusual position in studying and 
assessing his work. Ramsey was extremely productive, and his work always contains 
remarkable insight and originality. But his time was short, and probably none of his 
efforts approach the refinement of the treatments that we can imagine he would have 
produced, had he lived longer. 



Foundations of mathematics 

Though he later became a convert to finitism, Ramsey's well-known work on the foun- 
dations of mathematics explores and defends logicism, the view that mathematics is 
part of logic. Ramsey adopted the logic of Wittgenstein's Tractatus 1 and used its 
accounts of propositions and tautology to criticize and improve on the system of 
Principia Mathematica (PM). Russell and Whitehead used the ramified theory of types 
to avoid a variety of potential contradictions, including the paradox found by Russell 
that undermined the previous logicist theory of Frege (see FREGE and RUSSELL). PM's 
ramified type theory involved a double hierarchy of (1) types of classes and proposi- 
tional functions, and (2) orders of prepositional functions within each type. Both hier- 
archies were motivated by, and according to Ramsey sloppily deduced from, a vicious 
circle principle: its idea is that classes may not include themselves as members, nor can 
propositional functions meaningfully apply to, or quantify over, themselves. With its 
arrangement of classes into types according to their membership, the system of 
PM rejected the paradoxical set of all sets not members of themselves. With its insist- 
ence that meaningful propositional functions be defined to apply only to functions of 
lower order, the system avoided further contradictions, such as Richard's and Grelling's 

Ramsey criticized PM for recognizing only an impoverished range of classes, 
namely those definable through its comprehension axiom, but he had no objection to 
the type hierarchy it imposed on classes. He thought the ramified hierarchy of 
orders among functions of a given type was flawed in two ways, however. It missed 
the real nature of the paradoxes it sought to defuse, and it forced into the system of PM 
a nonlogical axiom, thereby undermining the logicist project. The axiom in question 
was the axiom of reducibility, asserting that for any propositional function of higher 
order there is an extensionally equivalent function of lowest order. With ramified 
types the axiom is needed, for instance, to guarantee that upper bounds of sets of real 
numbers will themselves be real numbers, rather than distinct entities with 
defining characteristics having a higher order than do the defining characteristics of 
the reals. Without the axiom PM could not capture important parts of mathematics 
(calculus and analysis, for example), but with it, Ramsey argued, PM made use of 
an axiom lacking logical necessity, and so failed its attempt to ground mathematics in 
logic alone. 

Ramsey cites Peano for noticing that Richard's paradox is linguistic rather 
than mathematical, but he is the one remembered for drawing a general distinction 
between logical and semantic paradoxes. The contradiction threatened by Russell's 
set would occur within a mathematical system, but not so with many other paradoxes. 
The liar paradox, Richard's paradox and Grelling's paradox (which Ramsey attributes 
to Weyl) "occur not in mathematics, but in thinking about mathematics . . . [they] 
could not be constructed without introducing the relation of words to their 
meaning or some equivalent" (1990: 184, 200). PM made the mistake of treating 
the semantic paradoxes as logical ones, bringing in ramified type theory, and with it, 
the reducibility axiom. Ramsey reconceived propositions and logical necessity along the 
lines of the Tractatus, and he provided a revised, simpler theory of types, now applying 
to the various symbolic expressions of propositions, for which the axiom of reducibility 



was not needed. "For me propositions in themselves have no orders; they are just 
different truth-functions of atomic propositions - a definite totality, depending only on 
what atomic propositions there are. Orders and illegitimate totalities only come in with 
the symbols we use to symbolize the facts in variously complicated ways" (1990: 

The logicist project was doomed by work done after Ramsey's death, but it seems 
Ramsey himself abandoned it before then. He had joked of preserving mathematics 
from the "Bolshevik menace of Brouwer and Weyl," but he later made extensive notes 
on intuitionist mathematics, and Braithwaite reports that he was converted to that 
view near the end of his life (1931: xii, 1990: 219, 1991a: 197-220). 

Belief and truth 

What is truth? For Ramsey, who takes truth to be first a property of beliefs and judg- 
ments, and only derivatively a property of sentences, this becomes the question, What 
is it for a belief to be true? His answer is that a belief is true when it is a belief thatp, 
and p. He regarded this as entirely obvious; the difficult part in analyzing the truth of 
belief is not with the concept of truth, but with the analysis of what it is to believe that 
p. To say that it is true that the earth is round amounts to saying that if anyone were 
to believe that the earth is round, their belief would be true. Which is to say "no more 
than that the earth has the quality you think it has when you think it is round, i.e. that 
the earth is round" (1990: 38-9, 1991b: 7-13). To say, "Everything he believes is true" 
amounts to no more than "For all p, if he believes p, then p." The latter sounds odd in 
ordinary discourse; our grammatical habits demand a verb and push us toward tacking 
an "is true" on to the end of the latter sentence, which then would hardly be enlight- 
ening. But the verb and the clause are unnecessary; a verb is already present in any of 
the beliefs he may have. 

Ramsey regarded his account of truth as a qualified correspondence theory, 
but it has come to be known as a redundancy theory. As he made clear, if in order to 
give an account of what it is to believe that p, he were forced to rely on the concept 
of truth, then not much progress would have been made. So what accounts for the 
contents of beliefs, judgments, and assertions? Ramsey sketched what he regarded as 
a pragmatist account, influenced by Russell and Peirce, that looks to the causal 
properties of our mental states. The primary target of his account is occurrent 
linguistically expressed belief, consciously asserted or denied; the contents of disposi- 
tional beliefs are derivative from the contents of occurrent thoughts. 2 The relevant 
mental states are silent or spoken linguistic utterances accompanied by feelings 
of assent or denial, and their contents are given by causal properties they bear to 
other mental states and to the world. Beyond describing the general picture, Ramsey 
devoted most of his attention to explaining how such causal properties exhibit 
patterns corresponding to the logical structures of the propositional contents born 
by the states. The key point is that his account of belief generates his account of 
meaning, rather than the other way around. "The essence of pragmatism I take to 
be this, that the meaning of a sentence is to be defined by reference to the actions 
to which asserting it would lead, or, more vaguely still, by its possible causes and effects" 
(1990: 51). 



Reasonable belief, probability, and knowledge 

"Truth and Probability" was written in 1926 and published posthumously (1931). 3 It 
is best known for its treatment of partial belief and subjective probability. Beyond this, 
however, Ramsey presented a view of logic construed as the science of rational thought, 
which he divided into two parts, the logic of consistency and the logic of truth. 4 

At the core of the views now known as probabilism, Bayesianism (in epistemology 
or decision theory), and subjectivism (about probability) is the idea of partial belief, and 
there is still no better introduction to it than the third section of "Truth and 
Probability." We hold some of our beliefs more strongly than we hold others. What is it 
that varies among the beliefs held with different strengths? Different beliefs will have 
different contents, of course, and sometimes the content of a belief may be about the 
strength of a belief, as when I believe that you strongly believe thatp, or doubt that q. 
I might have a view about the strength of my own belief, but that view (a second-order 
belief, we might say) is distinct from its subject (my belief thatp, with whatever strength 
it has). So if it is not in the content of a belief, where and what is the strength of a 
belief? And if we can answer that, can we go on to make systematic sense of the degrees 
of strength that beliefs can and do have? Perhaps we can do little better than to say that 
on different occasions we are certain, or confident, or think that maybe . . . , and so 
on. But Ramsey's answer to the first problem (What is strength of belief?) illuminates 
the point of distinguishing different degrees of strength, and it makes room for a range 
of strengths richer and more systematic than is the range indicated by our ordinary 

The essential idea is that the most fruitful way to think about the degrees of strength 
to which we hold beliefs (degrees of belief, for short), is to attend to the ways beliefs guide 
us in our choosing and acting. A degree of belief is a causal property of it, "which we 
can express vaguely as the extent to which we are prepared to act on it" (1990: 65). 
Ramsey proposed, first, that there is a univocal way in which a proposition p enters into 
a person's deliberations, so that it makes sense to speak of the degree to which she holds 
p.' If we are willing to accept that, then we can look to a person's deliberations and 
potential choices for indications of p's influence on them, i.e. for indications of her 
degree of belief in p. In doing this, we are exploring a measurement problem, and 
Ramsey was well aware of both the theoretical difficulties and the practical complica- 
tions that accompany a solution to it. On both scores, he drew an analogy to measure- 
ment in physical science. On the theoretical side, just as the length of a time interval 
between two events depends in relativistic physics on exactly how it is measured, so may 
the influence of a belief on choice. So just as when we use the idea of time intervals, 
when we use the idea of degrees of belief we should keep track of how we propose 
to measure them, but "for many purposes we can assume that the alternative ways of 
measuring it lead to [approximately] the same result" (1990: 63; see also p. 68). On the 
practical side, the intertwining of different physical influences and the disturbances 
introduced by measurement processes do not undermine all our attempts to understand 
and quantify physical phenomena. So it is at least not obvious that similar practical 
complications in studying belief and choice will defeat our attempts to do so. 

Ramsey first gave a familiar, if piecemeal, account. As he put it, "The old-established 
way of measuring a person's belief is to propose a bet, and see what are the lowest odds 



which he will accept. This method I regard as fundamentally sound; [though inexact 
and not general]" (1990: 68). The odds one will offer on a bet (ratios of its potential 
payoffs) indicate one's degree of belief. If I am willing to offer high odds as I defend p, 
my degree of belief in p is high; if I offer only low odds (demand high odds from my 
opponent) my degree of belief is low. A conditional degree of belief in p, given that q, is 
indicated by the odds I would place on a wager on p which only pays off in the cir- 
cumstance that q is true. Of course we undertake wagers infrequently compared to the 
frequency with which we have beliefs, and the betting scenario is quite artificial as a 
model for the variety of choices we make. Beliefs and their strengths are dispositions, 
though, and the measurements proposed via wagering are indicators of them, not the 
degrees of belief themselves. And in a wider sense, Ramsey said, all our lives we are 
betting: "Whenever we go to the station we are betting that a train will really run, and 
if we had not a sufficient degree of belief in this we should decline the bet and stay at 
home" (1990: 79). 

What degrees of belief ought we to have? The logic of consistency requires that 
degrees of belief obey the rules of probability. Apropos of the betting scenario, Ramsey 
stated - and seems to have understood better than most - what has become known as 
the Dutch book argument. Degrees of belief that violate probability would guide a 
person toward betting arrangements guaranteed to yield a loss (a certain loss, accord- 
ing to his own values) and Ramsey took this to be an indication of inconsistency in the 
partial beliefs. 

The most remarkable part of Ramsey's treatment of partial belief is his generaliza- 
tion of the betting scenario used so far. Generalize the idea of a bet on p to a gamble 
having the form, a if p, ft if not, where a and ft represent states of the world bearing 
value and obtaining according to whether or notp does. Some gambles will be favored 
over others, depending upon the values of a and ft, and on p. Ramsey showed that if a 
person's preferences among a rich set of these gambles is well arranged, according to 
stated principles (transitivity, for example, is one), then there are non-arbitrary mea- 
surements of all the values and of the probabilities of all the propositions p. These mea- 
surements are attributable to that person's values and beliefs, and can be taken to be 
the values and degrees of belief that guide his choices. This result is an early forerun- 
ner of similar demonstrations given by many later economists and philosophers, results 
that are usually taken as foundations for utility theory or decision theory. Ramsey 
emphasized the importance of the result for an account of partial belief, though this 
theory yields both. Important later theories that follow him in this are Savage's and 
Jeffrey's. 6 

Beyond the dictates of consistency, what degrees of belief are reasonable? With 
repeated acknowledgment of Peirce's influence, Ramsey conceived of the logic of truth 
along pragmatist lines. The best approach is to ask about the reasonability of the habits 
by which we arrive at and hold our beliefs. Always fully believe the truth is not bad advice, 
but it is not a very useful recommendation either. Nor is it useful as a standard for the 
sort of general habits open to humans, habits that so often yield partial beliefs rather 
than certainties. A more appropriate standard judges the habits according to how 
closely their partial beliefs correspond to the rate at which the beliefs are true. That is, 
the habit should yield a partial belief whose strength corresponds to the frequency with 
which relevantly similar beliefs are true. Ramsey used an illustration involving a belief 



about the wholesomeness of yellow toadstools; we can use the beliefs of weather fore- 
casters. A forecaster does well when it rains 70 percent of the time in which her degree 
of belief is 7/10. Ramsey's work on the consistency of partial belief is well known. It is 
worth emphasizing that he attached as much importance to this second standard 
calling for alignment of degrees of belief with frequencies of truth (and, further, that 
he was aware of the complexities of developing it, e.g., in identifying habits and speci- 
fying what classes of cases are relevant). To return to the action-guiding nature of 
partial belief: 

[B]elief of degree m/n is the sort of belief which leads to the action which would be best if 
repeated n times in m of which the proposition is true. ... It is this connection between 
partial belief and frequency, which enables us to use the calculus of frequencies as a cal- 
culus of consistent partial belief. And in a sense we may say that the two interpretations 
are the objective and subjective aspects of the same inner meaning, just as formal logic 
can be interpreted objectively as a body of tautology and subjectively as the laws of con- 
sistent thought. (1990: 84) 

Ramsey said less about knowledge than about reasonable belief. He regarded knowl- 
edge as true, certain belief produced by a reliable process, and so it appears he consid- 
ered it the extreme case of fully held true belief, backed up by a process that tends to 
produce such beliefs. This is a natural extension of his suggestion for evaluating our 
habits of belief, especially if the required reliability matches the strength of the belief 
(certainty). Though his most straightforward and explicit statements demand certainty, 
they are accompanied by discussions of fallibilism and by further remarks that soften 
the demand to near-certainty, practical certainty, or conviction "just a minute fraction 
short" of certainty (1990: 110—11, 1991b: 62-4). He agrees with Russell's view that 
"all our knowledge is infected with some degree of doubt," and in a paragraph on 
Moore's paradox and the paradox of the preface, he shifts from talk of certainty to talk 
of being nearly certain. It is likely a mistake to overanalyze his brief unpublished 
remarks, and it is difficult to determine how closely Ramsey's account of knowledge is 
tied to his account of reasonable belief. It is clear, though, that he endorsed a reliable- 
process account of knowledge, and he is remembered by contemporary epistemologists 
for doing so. 

Laws, causality, and theories 

In the last two years of his life Ramsey worked seriously on causality, laws of nature, 
and the formal structure of scientific theories. The several papers on these topics are 
not so finished as his earlier work, and in places indicate rapidly evolving views. This is 
also the period in which he was at work on his book on truth and moving away from 
logicism to an intuitionist view of mathematics. 

Ramsey's view of causality was not very distant from Hume's. In his brief 1928 
paper on law and causality (1990: 142-3), he suggests that the difference between uni- 
versal of law and universals of fact (between lawlike and accidental generalizations, 
we might say) lies in their distinct roles in our system of knowledge. If we knew every- 
thing and organized our knowledge in a deductive system that strove for simplicity, the 



general axioms of the system would be the fundamental laws of nature. They and the 
generalizations derivable from them without reference to facts of existence are the 
"statements of causal implication." And though really we do not know everything, we 
do tend to organize our knowledge in a deductive system, regard its axioms as laws, and 
regard as undiscovered laws the future axioms we expect to arise as we learn more. 
Ramsey soon revised this view, but its influence persists in the work of more recent 
philosophers, notably in David Lewis's best-system account of laws 7 (see LEWIS). 

Ramsey's revised treatment is in the 1929 paper, "General Propositions and 
Causality." The view there is that causal laws do not get their force by being simple fun- 
damental generalizations in an axiomatic summary of our knowledge. Their causal 
force lies in our trusting them as guides in our inferences about particular events. 
Causal generalizations "are not judgments but rules for judging 'If I meet a <|>, I shall 
regard it as a \|/.' This cannot be negated but it can be disagreed with by one who does 
not adopt it." An assertion of a causal law is an assertion not of a proposition, but of 
a formula from which we derive propositions about particular events. Its causal char- 
acter lies in the temporal ordering of the events about which it licenses our judgments 
(\|/ does not precede <|>). The special importance we attach to judgments with that order- 
ing is traceable to the importance of forward-looking judgments in our thinking about 
the influence our actions may have on the world. 8 In the course of discussing 
conditionals in this paper, Ramsey suggests that the acceptability of a conditional goes 
by the acceptability of its consequent after the antecedent is hypothetically added to 
one's beliefs: 

If two people are arguing "If p will q?" and are both in doubt as to p, they are adding p 
hypothetically to their stock of knowledge and arguing on that basis about q; . . . We can 
say that they are fixing their degrees of belief in q given p . . . (1990: 155) 

In contemporary work on conditionals this idea has become widely known as the 
Ramsey test for the acceptability of a conditional. 

In another 1929 paper, Ramsey addresses the formal structure of scientific 
theories ("Theories," in Ramsey 1990). He is particularly interested in the question of 
the content of theoretical assertions, and how such content is related to the 
observational assertions on which the theory is built. One idea for demonstrating the 
dispensability of theoretical terms is to show that they are explicitly definable in 
terms of the observational assertions, and further, that the definitions can be inverted 
so that anything we express in theoretical language can also be expressed without 
recourse to the theoretical terms. Ramsey works all this out for a very simple example; 
even for that example, the results are complex and extremely cumbersome. Worse, 
as he points out, the method of explicit definitions creates the problem that additional 
data not seriously inconsistent with the theory will nevertheless falsify the theory 
unless adjustments are made to the meanings of its terms. Is there another alternative? 
Ramsey offers one. Conjoin all the sentences of the (first-order) theory, replace 
the occurrences of each distinct theoretical term with a second-order variable, and 
introduce for each distinct variable a second-order existential quantifier that binds 
its occurrences. The result is now known as the Ramsey sentence of the theory. It 
contains only observational terms, is entailed by the first-order theory, and it entails the 



same particular observation sentences as the first-order theory. This device has since 
been used by Hempel, Carnap, and many others in treatments of the content and 
meaning of theories, whether the concern is, as above, with the observational content 
of scientific theories or with, for example, the relation between mental and neuro- 
physical theories. 


1 L. Wittgenstein, Tractatus Logico-Philosophicus, London: Routledge, 1922. Ramsey con- 
tributed to the 1922 English translation and made criticisms that led to changes in the 1933 
second edition. Wittgenstein mentions Ramsey's influence in the preface to his Philosophical 

2 Dispositional linguistic belief, that is; Ramsey also entertained the idea that a chicken has 
beliefs, to be understood as relations between its potential behavior and circumstances in 
which the behavior is appropriate. This resembles the analysis he gives of dispositional 
strengths of belief (see below); Ramsey 1990: 40. 

3 Several later notes expressing afterthoughts and further ideas are in Ramsey 1931, 1991a. 

4 The organization of "Truth and Probability" was the starting point for a planned book. Much 
of the unfinished manuscript for the book has been published as Ramsey 1991b. 

5 Notice that the idea of a single degree of belief is for a person, at a time, among a collection 
of other beliefs. A degree of belief may very well change for a variety of reasons, including 
observations, the passage of time, or changes in other beliefs. 

6 L. J. Savage, The Foundations of Statistics [1954], 2nd edn., New York: Dover, 1972; 
R. C. Jeffrey, The Logic of Decision [1965], 2nd edn., Chicago: University of Chicago Press, 

7 D. Lewis, Philosophical Papers, vol. II, Oxford: Oxford University Press, 1986. Lewis acknowl- 
edges "following the lead of (a short temporal segment of) Ramsey," p. xi. 

8 Ramsey calls these rules for judging variable hypotheticals, of which causal laws are an 
important kind. Ramsey 1990: 149, 157-9. 


Ramsey's best-known and most influential work is found in the collections of 1931, 1978, and 
1990; their philosophical contents are very similar. His surviving papers are in the Ramsey 
Collection housed at the University of Pittsburgh's Hillman Library. A fascinating group of these 
papers, almost all otherwise unpublished, appear in the collection dated 1991a. A substantial 
manuscript on truth, also from the Ramsey Collection, was published in 1991b. The contents of 
Mellor (1980) are more inspired by Ramsey than centered on his work. Bibliographies of 
Ramsey's work appear in Ramsey 1931, 1978, 1990, and in Sahlin 1990; all of the books listed 
below contain interesting introductions or prefaces. 

Works by Ramsey 

1931: The Foundations of Mathematics and Other Logical Essays, ed. R. B. Braithwaite, London: 

Routledge and Kegan Paul. 
1978: Foundations, ed. D. H. Mellor, Atlantic Highlands, NJ: Humanities Press. 
1990: Philosophical Papers, ed. D. H. Mellor, Cambridge: Cambridge University Press. 
1991a: Notes on Philosophy, Probability, and Mathematics, ed. M. C. Galavotti, Naples: Bibliopolis. 
1991b: On Truth, ed. N. Rescher and U Majer, Dordrecht: Kluwer. 



Works by other authors 

Mellor, D. H. (ed.) (1980) Prospects for Pragmatism: Essays in Memory of F. P. Ramsey, 

Cambridge: Cambridge University Press. 
Sahlin, N. (1990) The Philosophy of P. P. Ramsey, Cambridge: Cambridge University 



Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 


Carl G. Hempel (1905-1997) 



Carl Gustav Hempel was one of a group of philosophers from Central Europe who 
emigrated to the United States in the 1930s and who profoundly modified the 
character of American philosophy. Together with Rudolf Carnap, Ernest Nagel, and 
Hans Reichenbach, Hempel was central to the transition from logical positivism to 
logical empiricism. His writings not only set the agenda for philosophy of science in the 
middle decades of the twentieth century but also continue to shape this important field 
of philosophy. 

Educated in Berlin, Hempel was influenced by early twentieth-century attempts to 
apply the concepts and techniques of mathematical logic to the empirical sciences, pio- 
neered by Reichenbach in Germany and Carnap and his co-workers in Vienna. The 
Vienna Circle had hoped to diagnose most traditional philosophical discussions as treat- 
ments of pseudo-problems, by formulating and applying a precise criterion of cognitive 
significance. The envisaged criterion was intended to pick out, as meaningful, state- 
ments of logic and mathematics conceived as analytic truths, and those non-analytic 
statements that admit of empirical test. The sciences, paradigmatically the physical sci- 
ences, were to count as meaningful because they satisfied the latter condition, whereas 
the statements of traditional philosophy were to be exposed as neither analytic truths 
nor susceptible to empirical test, and consequently devoid of cognitive significance. 

During the 1930s, logical positivists made successive attempts to make the criterion 
of cognitive significance sufficiently sharp to perform the planned surgery. Hempel par- 
ticipated actively in these discussions, examining various proposals and identifying dif- 
ficulties. His efforts culminated in synthetic essays that argued for the impossibility of 
the positivist project and drew the blueprint for logical empiricism, the philosophical 
structure within which he would build substantive positions. 

Cognitive significance 

Hempel never wavered in his commitment to the idea that "the general intent of the 
empiricist criterion of meaning is basically sound" (1965: 102), that is, that mean- 
ingful empirical statements are those that admit of experiential test. He believed, 



however, that attempts to state precise logical criteria of empirical significance 
encounter systematic difficulties. If we insist that empirical statements are those 
that admit of conclusive verification, then the criterion will debar universal statements 
(including the laws of the various sciences); to propose that empirical statements 
must allow conclusive falsification is equally hopeless, since this would deny meaning- 
fulness to existential claims. Nor can we propose to formulate a criterion of meaning- 
fulness by requiring definability of all terms in some language whose nonlogical 
vocabulary can be learned in application to observational entities and properties, for 
the special terms that play so fruitful a role in the physical sciences, expressions like 
"state function" and "covalent bond" cannot be given explicit definitions. Hempel con- 
cludes that the requirement of meaningfulness must imitate the way in which scien- 
tists extend the resources of everyday language, to wit by systematically connecting 
sentences in which unfamiliar expressions appear with sentences whose vocabulary is 

My reconstruction condenses a wealth of subtle points that Hempel articulates with 
characteristic lucidity, but it brings out two important aspects of the critique of formal 
criteria of cognitive significance. First, Hempel is not simply concerned with a philo- 
sophical enterprise that intends to transcend what are taken to be sterile disputes, but, 
in addition, to provide an assessment of movements in the natural and social sciences 
that call for empirical conditions for the application of bits of theoretical vocabulary. 
He offers a definitive rebuttal to the operationalist demand that newly-introduced 
terms must be associated with an experiential procedure for applying them (see Hempel 
1965: ch. 5). Second, that rebuttal depends on taking the unit of appraisal to be 
"sentences forming a theoretical system" (1965: 117). Explicitly acknowledging that 
this makes cognitive significance, "a matter of degree" (ibid.), he offers four main ways 
in which theoretical systems should be evaluated: by scrutinizing the connections 
among theoretical terms and the connections to statements couched in observational 
vocabulary; by considering the explanatory and predictive power of the system; by 
appraising the simplicity of the system; and by assessing the degree of confirmation 
by empirical evidence. 

For the positivist eager to prick the pretensions of traditional metaphysics or the 
operationalist concerned that some area of science should be rigorously developed, this 
concluding catalogue must come as a disappointment. In place of a clear criterion that 
can put an end to disputes, it appears that these fuzzy virtues of theoretical systems will 
be hard to identify and that we are fated to continue the wrangles of the past. Hempel's 
reply would be twofold. First, he would point out that the weapons for which positivists 
and operationalists yearn are simply not to be had. Second, he would insist that his list 
of modes of appraisal may be the end of a reflection on the fate of positivism, but that 
it is only the beginning of a serious philosophy of science. The task for logical empiri- 
cism is to say, as precisely as possible, what kinds of linguistic connections are present 
in virtuous theoretical systems, what makes for explanatory and predictive power, 
what counts as simplicity, and how empirical statements are confirmed by the results 
of experiential tests. 

For Hempel, then, there are four main problems of the philosophy of science. 
The last problem, the issue of empirical confirmation, is also central to the theory of 
knowledge, and, since ontological questions depend on obtaining a clear view of the 



kinds of entities to which the sciences commit us, clarification of theoretical structure 
is pertinent to discussions in metaphysics. Throughout his career, Hempel made major 
contributions to three of the four problems he highlighted: he offered a theory of quali- 
tative confirmation, provided magisterial discussion of the fruitful use of theoretical 
vocabulary, and delineated an account of scientific explanation against which all sub- 
sequent treatments of this topic must be measured. Other logical empiricists, particu- 
larly Reichenbach (1938) and Nelson Goodman (1949), took up the problem of 
understanding simplicity, although this topic never achieved the same prominence 
within logical empiricism as the other issues. 

A philosopher's life work rarely conforms to a neat structure, and Hempel's is no 
exception. During the course of his career, he wrote important essays on the character 
of mathematics and the relations between mathematics and the natural sciences 
(1945a, 1945b). He also retained a strong interest in the properties of systems of clas- 
sification, from his early attempt to apply logical notions in taxonomy (Hempel and 
Oppenheim 1936) to various attempts to evaluate proposed classificatory schemes in 
psychiatry and in the social sciences (1965: chs 6 and 7). Despite the undeniable influ- 
ence that these studies have had, Hempel's attacks on the problems of confirmation, 
theory-structure, and explanation are, I believe, his most enduring accomplishments. 
The following sections will consider them in order of ascending importance. 

Qualitative confirmation 

Suppose that h is a hypothesis in whose truth-value we are interested, and that e is a 
statement that reports the result of some empirical test. The general problem of con- 
firmation is to understand the nature of the relation of evidential support between h 
and e. This general problem encompasses several specific questions: we might ask the 
degree to which e would support h (the quantitative problem): alternatively, we might 
inquire after the conditions under which e would support h at all (the qualitative 
problem); an intermediate question probes the conditions under which e supports h 
more than e' supports h' (the comparative problem; note that e might be the same as e' 
or h identical with h'). In his extensive investigations of inductive logic, Carnap focused 
on the quantitative problem, considering formal languages adequate for the formula- 
tion of fragments of science, and attempting to define, for a wide class of statements h 
and e within these languages, the degree to which e would confirm h. By contrast, 
Hempel took the qualitative problem to be more fundamental, and endeavored to specify 
conditions under which singular statements (conceived as ascribing properties to 
objects) would confirm, disconfirm, or be neutral to, a hypothesis (characteristically 
thought of as a lawlike generalization). 

Hempel's treatment is noteworthy not just for his positive proposal but for his 
disclosure of interesting difficulties with apparently plausible ideas. Suppose that 
the hypothesis of interest is the generalization that all ravens are black, formalized 
as (x)(Rx zd Bx). It is very natural to believe that the generalization is supported by 
observing black ravens. So we might arrive at the general proposal of confirmation by 
instances: the hypothesis (x)(Rx zd Bx) is confirmed by any sentence Ra&Ba. As Hempel 
points out, the manner in which we formulate a hypothesis should make no difference 
to the class of statements that confirm it. Thus, if h and h' are logically equivalent any 



e that confirms h should confirm h', and conversely. By elementary logic, (x)(Rx z> Bx) 
is logically equivalent to (x)(-Bx z> -Rx). The thesis of instance confirmation now tells 
us that the latter is confirmed by -Ba&-Ra, which, by the principle about logical equiv- 
alence, must also confirm (x)(Rx z> Bx). Reverting to our interpretation of the nonlog- 
ical vocabulary, we discover that the generalization "All ravens are black" is confirmed 
by statements that tell us that a particular object is neither black, nor a raven. 
Apparently, learning that the rightmost shoe in my closet is white would support an 
ornithological generalization! 

The "paradox of the ravens" has inspired a large subsequent literature. Hempel's 
own diagnosis was that there is no genuine paradox, and that any sense of surprise 
stems from "misguided intuitions" (1965: 20). Whether or not this is so, there is no 
doubt that Hempel's further investigations disclose severe problems in natural 
conceptions. Many philosophers, and scientists reflecting on methodological issues, 
have accepted the fundamental idea of hypothetico-deductivism, to wit that hypothe- 
ses are confirmed when their consequences are found to be correct. The most straight- 
forward way to formulate that idea is as the Converse Consequence Condition: if e is a 
consequence of h, then e confirms h. Unfortunately, that condition, coupled with a 
requirement that is hard to resist, generates the conclusion that any statement will 
confirm any hypothesis. The further requirement is the Entailment Condition: evidence 
statements confirm the logical consequences of the hypotheses they confirm. Given any 
statement e, e is a consequence of h&e (whatever h may be); so by the Converse 
Consequence Condition e confirms h&e; h is a consequence of h&e; hence by the 
Entailment Condition e confirms h. 

Hempel's own account of qualitative confirmation avoided this difficulty by aban- 
doning the Converse Consequence Condition. Instead, he proposed that direct 
confirmation results when an evidence statement entails a restricted version of the 
hypothesis, effectively what the hypothesis would say if there just existed the individ- 
uals mentioned in the evidence statement. More exactly, suppose that the evidence 
statement e contains the names of exactly the individuals a v . . . , a„; then e directly con- 
firms (x){Rx z> Bx) just in case e entails each of the statements R«j z> Bdj Ra n zd 

Ba n . The general notion of confirmation is obtained by proposing that e confirms h 
just in case h is entailed by a class of sentences each member of which is directly 
confirmed by e. 

Hempel's approach is unable to account for the confirmation of sentences contain- 
ing theoretical vocabulary by statements formulated in more basic terms. Clark 
Glymour attempted to extend the Hempelian approach to offer an account of qualita- 
tive confirmation (or of relevant evidence) that would address this difficulty (Glymour 
1980), and his proposal has given rise to extensive subsequent discussion. Hempel 
himself became convinced that the general approach could not succeed, on the grounds 
that Nelson Goodman's "new riddle of induction" showed the inadequacy of any purely 
syntactical analysis of qualitative confirmation (see GOODMAN). Arguing that not all 
universal generalizations are supported by their instances, Goodman (1955) exposed 
the difficulties of distinguishing those generalizations that can be confirmed in this way 
from those that cannot. 

Ironically, much contemporary thinking about confirmation diverges from Hempel's 
treatment at a very early stage. The most influential more recent proposal is 



Bayesianism, a position that descends from Carnap's investigations in inductive logic 
and that attempts to understand how degrees of confirmation of hypotheses adjust 
in the light of evidence. For Bayesians, the problem of quantitative confirmation is 
primary and the solution to the qualitative problem is generated in a trivial fashion from 
a solution to the quantitative problem. To solve the latter, we need to be able to assess 
the value of the probability of the hypothesis given the evidence, Pr(h\e); assuming that 
that can be done, we can say that e confirms h (or confirms h relative to background 
information B) just in case Pr(h\e) > Pr(h) (or Pr(h\e&B) > Pr(h\B)). Hempel's discussions 
of confirmation remain of interest not so much because of his positive proposal about 
qualitative confirmation as for his careful exposure of difficulties in intuitively attrac- 
tive ways of thinking about confirmation, and for his recognition of constraints on any 
adequate solution. 


Logical empiricism began with the conviction that the tools of logic developed by Frege, 
Russell, and their successors could be used to make clear and explicit the structure of 
scientific theories (see FREGE and RUSSELL), and, indeed, even in the 1920s, Reichenbach 
had offered an axiomatization of the special theory of relativity, intended to exhibit 
which parts of the theory were conventional stipulations and which made substantive 
empirical claims. Almost unselfconsciously, the logical empiricists took over the logi- 
cian's conception of a theory as a deductively closed set of sentences in some suitable 
formal language. Recognizing that an important - and, from the empiricist viewpoint, 
problematic - feature of the theories in physics and chemistry that most impressed 
them was the presence of special vocabulary that resists explicit definition in observa- 
tional terms, logical empiricists formulated a distinctive view about scientific theories. 
A scientific theory is a deductively closed set of statements in a first-order language 
whose nonlogical vocabulary divides into two subsets, the basic vocabulary (often 
understood as containing those terms whose application can be made on the basis of 
more or less direct observation) and the theoretical vocabulary (the remainder); the 
statements whose essential nonlogical vocabulary contains only theoretical terms are 
the theoretical postulates of the theory, while statements whose essential nonlogical 
vocabulary contains both theoretical and basic terms are the correspondence rules (a 
variety of other designations for this last class of statements appears in logical empiri- 
cist writings, but "correspondence rules" is the most popular locution); the function of 
the correspondence rules is to provide the theories with empirical content, and they (or 
a subset of them) are often conceived as providing an interpretation (or partial inter- 
pretation) of the theoretical vocabulary. 

This conception of scientific theories became fully explicit in the writings of Carnap 
(1956), Nagel (1962), and Hempel (1958, 1965: ch. 8), and, perhaps as a residue of 
the concerns about cognitive significance, each of these authors sought ways to char- 
acterize those theoretical contexts in which the introduction of theoretical vocabulary 
served important scientific purposes (see CARNAP). Hempel's discussions revolved 
around a family of questions. To what extent can the correspondence rules be viewed 
as functioning as definitions? Is it problematic to concede that the correspondence rules 
only offer partial interpretations of the theoretical vocabulary, or should this be seen 



as a symptom of the openendedness of scientific research? Can we eliminate the theo- 
retical vocabulary without any scientific loss? Is it reasonable to treat the theoretical 
terms as components of a formal apparatus for making experiential predictions, or 
should we suppose that those terms refer to entities and properties that underlie the 
observable phenomena? 

Hempel's discussion of these questions embodies a cautiously realistic attitude. He 
does not think that we can provide a full explicit definition of theoretical vocabulary in 
observational terms, not even for those relatively low-level parts of science that make 
use of dispositional concepts. Although he believes that correspondence rules are vehi- 
cles of partial interpretation, he suggests that we cannot neatly separate the parts of a 
theory that function to pin down the meanings of our terms from those that make 
genuinely empirical claims, a point in which he concurs with W. V. Quine's celebrated 
critique of the analytic/synthetic distinction (Quine 1953) (see QUINE). Partial inter- 
pretation, Hempel believes, has heuristic advantages, allowing us to introduce new cor- 
respondence rules as we extend the theory to cope with previously untreated 
phenomena. Further, to the extent that theoretical vocabulary is ineliminable, he holds 
that the evidence leading us to adopt the theory ought to incline us to accept its pos- 
tulates as true and thus to treat its theoretical vocabulary as referring to entities beyond 
the reach of ordinary observation. 

The crucial issue thus turns out to be whether or not there is a generally available 
method for eliminating theoretical vocabulary. Hempel approaches the problem by 
formulating a dilemma. Starting from the premise that the function of a theory is to 
"establish definite connections among observable phenomena" (1965: 186), he sug- 
gests that when such connections are established we do not need any detour through 
a theory, since the theory will imply a conditional statement, couched in the basic 
vocabulary, that asserts the connection. So, if the theoretical terms and principles serve 
their purpose, they can, in principle, be eliminated, and are thus unnecessary. If, on the 
other hand, the theoretical terms and principles do not establish the intended connec- 
tions, they do not serve their purpose, and are consequently unnecessary. This is "the 
theoretician's dilemma" (1965: ch. 8). 

Behind the dilemma stands a view of the use of theories in scientific practice. It is as 
though the scientist feeds some description of observables into the theoretical machin- 
ery, the gears turn, and the output is another statement about observables. The point 
may be to predict something (the output statement is one we didn't know before) or it 
may be to explain something (we already knew the output statement but didn't see why 
it was true). In either case, the theoretical statements function to license an inference 
from the input to the output, and thus must support the conditional statement, "If 
input then output." Why then can we not manage with the set of all such conditional 
statements corresponding to the transitions that the theory would license? 

A first response is that the resultant set would be extremely unwieldy, that the theory 
as actually presented provides a concise way of representing a disparate class of con- 
sequences. As Hempel and his colleagues clearly saw, however, a result due to the 
logician William Craig constructs a recursive procedure for generating the class of con- 
sequences, without stepping outside the basic vocabulary. (It is interesting to reflect that 
the significance attributed to Craig's theorem revealed the hold that the logician's con- 
ception of theories as recursively axiomatizable continued to exert on logical empiricist 



discussions of scientific theories.) fn the end, Hempel's response to this and to kindred 
suggestions for eliminating theoretical vocabulary draws on the proposal that the task 
of a scientific theory is not only to achieve deductive systematization of observable phe- 
nomena but also to provide inductive systematization, and we have no reason to believe 
that any of the elimination procedures will satisfy this further constraint (1965: 
214-15; note that in the case of the Craigian surrogate, Hempel is able to argue that 
the substitute will not achieve any inductive systematization). 

ft is at first sight ironic that the philosopher who contributed most to our under- 
standing of scientific explanation overlooked a relatively obvious point about proposals 
for eliminating theoretical terms: even though they might be able to mimic the predic- 
tive successes of genuine theories, it seems that they would incur severe explanatory 
losses. Whatever set of brute empirical rules we might devise for predicting the out- 
comes of bringing substances together in various proportions would fail to deliver the 
explanatory benefits we obtain from embedding empirical generalizations within the 
theoretical treatment of shell-filling and of ionic and covalent bonds. On deeper reflec- 
tion, however, we can see that appeal to explanatory power would have led Hempel into 
uncomfortable questions about the sufficiency of his preferred account of scientific 
explanation. For there is every reason to think that some of the procedures for elimi- 
nating theoretical terms would not just deliver singular conditional statements but 
generalizations from which such singular conditionals could be derived; because the 
generalizations would serve as the needed "covering laws" in Hempel's schemata for 
explanation, by the standards of the Hempelian account of explanation the explana- 
tory loss would be indiscernible. 

During the 1960s and 1970s, the account of theories favored by Hempel, Nagel, 
Reichenbach, and other logical empiricists acquired the name "the received view," and 
like most doctrines so designated came under vigorous attack (see Suppe 19 70 for thor- 
ough analysis). One principal difficulty, recognized by Hilary Putnam, was that the 
account conflated two distinctions, the distinction between observational and theoreti- 
cal terms and that between observable and unobservable things; as Putnam noted, some 
theoretical terms name observable things ("oscilloscope") and unobservable things can 
be picked out using observational terms ("people too little to see"). Putnam's observa- 
tions, and the discussions of reference that he and others initiated (Kripke 1971, 
Putnam 1973) undercut the old concern that scientists are simply unable to specify the 
referents of their theoretical vocabularies. 

A different line of objection attacked the idea that a theory is a linguistic item. 
Several authors (including Suppes 1967, van Fraassen 1980) drew inspiration from 
model theory rather than from the syntax of logical systems, proposing that theories 
are to be identified with families of models. Their critiques initiated a debate, as yet 
unresolved, about the correct analysis of the notion of a scientific theory, ft is perhaps 
worth recalling that both the newer "semantic conception of theories" and the older 
"syntactic account" (or "received view") are philosophical reconstructions of the prac- 
tices of scientists, and that the standards of adequacy for a reconstruction depend on 
what purposes - philosophical or scientific - one intends to achieve. The question 
"What is the real structure of a scientific theory?" may simply be a bad question, and, 
depending on our aims, we may draw on one or another of the proposed accounts (or 
on something completely different). To the extent that the syntactic conception con- 



tinues to be valuable in such enterprises, we can expect that Hempel's lucid and careful 
delineations of possibilities and constraints will remain pertinent. 


In the early decades of the twentieth century, the thought that one of the aims of the 
sciences is to provide explanations, traditionally popular, suffered a temporary eclipse. 
Thinking that appeals to the explanatory power of a theory reflected purely subjective 
judgments, scholars writing about science tended to concentrate on the criterion of 
predictive success (see, for one among many examples, Pearson). Hempel played the 
leading role in restoring the respectability of the concept of scientific explanation. From 
his earliest discussions of the topic, he insists on the objective character of scientific 
explanation (see 1965: 234 (originally written in 1942); also 1966: ch. 5). Taking up 
a theme already sounded by earlier empiricists (for example John Stuart Mill), and 
perhaps as old as Aristotle, he suggests that explaining a fact, state, or event consists in 
showing why that fact, event, or state could have been expected to occur, given the laws 
of nature. The key to explanation is nomic expectability 

Hempel proposed that an explanation is an argument whose conclusion is a state- 
ment describing the phenomenon to be explained (this statement is the explanandum) 
and whose premises (the explanans) include at least one law of nature. Although his 
early writings concentrated on cases in which the argument is deductive, he was 
explicit, from the beginning, that some explanations are non-deductive arguments. He 
also took considerable pains to point out that, as actually given, explanations may not 
take the ideal form he specified. So, for example, historians develop explanatory narra- 
tives that are far from complete arguments, and yet, Hempel contended, the explana- 
tory force of their work derives from the possibility of recognizing general laws of 
nature, which in combination with the claims they advance would yield a compelling 
argument for the explanandum. 

The Deductive-Nomological (D-N) model of explanation can be encapsulated in a 
schema: deductive explanatory arguments take the form 

w, . . . , u m 
Li, ■ . . , L n 


where the Q are statements reporting particular facts, the L : are laws of nature, whose 
presence is essential to the validity of the argument, and E is the explanandum. 
(Hempel's model does not require that there be any Cs, although there must be at least 
one L; his presentations sometimes identify E as a statement of particular fact, but he 
allows for explanations of this form whose conclusions are laws, including probabilis- 
tic laws.) To be a genuine explanation, the premises of an argument fitting the schema 
must all be true. If one or more of the premises is not true, then the argument counts 
as a potential explanation. 

In the 1940s, Hempel hoped to articulate the D-N model more precisely, and he pro- 
posed a formal explication of the notion of law and of deductive explanation (Hempel 
and Oppenheim 1948; see Hempel 1965: ch. 10). Unfortunately this attempt proved 
vulnerable to trivializing counterexamples, and, in any event, Goodman's explorations 



of laws, counterfactuals, and induction, convinced Hempel that no formal account of 
scientific laws could be given. Thus, throughout the 1950s and 1960s, his work on 
scientific explanation focused on showing how his preferred approach to explanation 
illuminated aspects of the natural and social sciences and how it could be extended 
to include non-deductive arguments. 

The latter task was complicated by an important disanalogy between deductive and 
inductive arguments. Adding extra premises to a deductively valid argument preserves 
validity, but the incorporation of new information into an argument that is inductively 
strong may not only undermine the argument but even support a contrary conclusion. 
So, to take one of Hempel's own examples, to claim that Jones is suffering from a strep- 
tococcal infection and that he is being treated with penicillin, together with the prob- 
abilistic law that 9 9 percent of those treated with penicillin recover from such infections 
confers high probability on the conclusion that Jones will recover, but if we now learn 
that this particular streptococcal infection is penicillin-resistant then we have strong 
reasons for thinking that Jones will not recover. 

Hempel's model of Inductive-Statistical Explanation (I-S) proposed that I-S explana- 
tions are arguments with true premises of the form: 

Pr(B\A) = r 



Here the double line indicates that the premises bestow on the conclusion the prob- 
ability r, which is supposed to be close to one. To block the problem of the "ambiguity 
of statistical explanation," Hempel imposes the "requirement of maximal specificity." 
If s is the conjunction of the premises of the explanation, and if k is a statement logi- 
cally equivalent to the set, K, of accepted sentences, then if s&k implies that c belongs 
to a subset A* of A, then s&k must also imply that Pr(B\A*) = r* where r*=r unless the 
conditional probability of B on A* is simply a matter of probability theory (as, for 
example, when A* is the null set). This intricate condition is intended to require that we 
always employ the most specific probabilistic information we have, and, as Hempel 
explicitly noted, it introduces an unwelcome relativization into the account of expla- 
nation, for, unlike D-N explanations, f-S arguments only qualify as explanations rela- 
tive to a particular state, K, of our knowledge. 

The Covering Law Model of Explanation, comprising the D-N and l-S models, was enor- 
mously influential, not only restoring the respectability of the concept of explanation 
but also sparking methodological discussions in the social sciences. The many-sided 
character of Hempel's lucid discussions, especially in the title essay of Aspects of 
Scientific Explanation, provides a model for philosophical exploration of an important 
metascientific concept. Nonetheless, for all the subtlety of his treatment, Hempel's 
account is no longer widely accepted among philosophers of science (although it con- 
tinues to be adopted in other philosophical debates and in the methodological reflec- 
tions of natural and social scientists). 

Some of the difficulties stem from problems with which Hempel struggled. The intri- 
cate requirement of maximum specificity proved inadequate to salvage the notion of 



I-S explanation, yielding the unwelcome consequence that bona fide inductive explana- 
tions turn out to be tacitly deductive (Coffa 1974). Another difficulty of the model of 
probabilistic explanation lies in the fact that we can apparently explain events that are 
unlikely to occur: even though it may be improbable that an atomic nucleus will 
undergo a particular sequence of decay, we can still, it seems, use quantum physics to 
explain the rare occurrences that do take that path. Even in the realm of purely deduc- 
tive explanation, there are formidable challenges. As Hempel himself noted, a deriva- 
tion of Boyle's Law from the conjunction of this law with Kepler's laws would satisfy 
the D-N schema, even though any such derivation is explanatorily worthless. How do 
we distinguish such arguments from the explanatory derivations of laws, for example 
the derivation of Kepler's laws within Newtonian gravitational theory? 

Perhaps the most severe problems came from a cluster of examples that showed how 
familiar asymmetries that occur in the context of causal judgments also affect our 
assessments of explanatory power. Suppose that a flagpole casts a shadow of a particu- 
lar length. Using the law of rectilinear propagation of light, together with facts about 
the height of the pole and the elevation of the sun, it is possible to derive the length of 
the shadow, a derivation that fits the D-N model. So far, so good, since that particular 
derivation seems genuinely explanatory. The trouble is that we can also work in the 
opposite direction. Given the propagation law, the elevation of the sun and the length 
of the shadow, we can derive the height of the pole, and this derivation fits the D-N 
schema equally well (Bromberger 1966). A natural response to examples like this is to 
declare that opaque objects produce (or cast) shadows and that shadows do not produce 
the associated objects, so that there is a causal asymmetry unrepresented in the 
Hempelian schema. 

Hempel's scattered remarks about the connection between explanation and causa- 
tion present a clear picture of his position. Influenced by Humean worries about the 
notion of causation, he holds that our understanding of causal relations is grounded 
in our ability to subsume phenomena under lawlike regularities. The concept of expla- 
nation is prior to that of causation, in that a claim that c caused e is always derivative 
from the thought that the occurrence of e would be properly explained by an argument 
in which a description of c figured among the premises, an argument satisfying the 
covering-law model. Hence, Hempel cannot appeal to causal asymmetries to reformu- 
late his account of explanation, and, on the few occasions on which he confronts exam- 
ples that embody such asymmetries, he argues strenuously that our intuitive responses 
to these cases should not be trusted (for example, 1965: 352-3). With the articulation 
of a family of instances like that of the shadow-casting flagpole, attempts to deny dif- 
ferences in explanatory worth came to appear ever more desperate, and most philoso- 
phers of science (including Hempel himself) have concluded that the problem of 
explanatory asymmetry cannot be dismissed as illusory. 

For about a quarter of a century Hempel's account of scientific explanation almost 
achieved philosophical consensus. Since it succumbed to a host of problems and criti- 
cisms no successor approach has garnered similar support. Inspired by the problem of 
asymmetry, several philosophers have offered accounts that invoke the concept of cau- 
sation (see, for example, Humphreys 1989, and Salmon 1984, 1998). Others have tried 
to preserve the main features of Hempel's account by developing an idea that received 



passing attention in his own writings, and suggesting that explanation consists in the 
unification of the phenomena (Friedman 19 74, Kite her 1981). Yet others have con- 
tended that explanation is an activity whose crucial properties vary with context 
(Achinstein 1983, van Fraassen 1980: ch. 5). All the existing accounts face major 
obstacles (often gleefully noted by the partisans of rival accounts). If there is a con- 
sensus, its central tendency is that, while Hempel's covering-law model is inadequate, 
it is exemplary in demonstrating the range, rigor, and clarity that any satisfactory 
theory of explanation should strive for. 

Hempel's legacies 

In the past decades, logical empiricism has been criticized for various shortcomings: 
neglect of the historical development of science (Kuhn 1962/1970), overemphasis on 
the search for lawlike regularities in nature (Cartwright 1983, 1999), and failure to 
appreciate the autonomy of experimental practice (Galison 1987, Flacking 1983). At 
the same time, many philosophers have proposed that Hempel and his co-workers 
adopted an unnecessarily restrictive view of the formal resources on which accounts 
of confirmation, theories, and explanation might draw. Despite these complaints, phi- 
losophy of science continues to pursue the agenda that Hempel so lucidly articulated, 
and, if the set of questions has been enlarged and the Hempelian answers are no longer 
widely accepted, it would be foolhardy to tackle these problems without thorough 
awareness of Hempel's many insights. 

No essay on Carl ("Peter") Hempel would be complete without some recognition of 
his extraordinary pedagogical influence. Not only was he the author of one of the great 
introductions to any field of philosophy (Hempel 1966), but, through lectures and semi- 
nars, he was an inspiration to generations of undergraduates, graduate students, and 
younger philosophers. Those who knew him saw, again and again, a rare combination 
of high scholarly integrity and personal kindness, acute intelligence and gentleness, and 
his daily actions reminded those around him that philosophy began with the desire for 
wisdom and for understanding the good. In his life, as well as in his work, Hempel was a 
true philosopher. 


Works by Hempel 

1936 (with Oppenheim, P.): Der Typusbegriff im Lichte der neuen Logik, Leiden: Sitjhoff. 
1945a: "Geometry and Empirical Science," American Mathematical Monthly 52, pp. 7-1 7. 
1945b: "On the Nature of Mathematical Truth," American Mathematical Monthly 52, 

pp. 543-56. 
1948 (with Oppenheim, P.): "Studies in the Logic of Confirmation," Philosophy of Science 15, pp. 

13 5-75. (Reprinted in Hempel 1965, ch. 10.) 
1958: "The Theoretician's Dilemma," in Minnesota Studies in the Philosophy of Science II, ed. H. 

Feigl, M. Scriven, and G. Maxwell, Minneapolis: University of Minnesota Press, pp. 3 7-98. 

(Reprinted in Hempel 1965, ch. 8.) 
1965: Aspects of Scientific Explanation, New York: Free Press. 
1966: Philosophy of Natural Science, Englewood Cliffs, NJ: Prentice-Hall. 



Works by other authors 

Achinstein, P. (1983) The Nature of Explanation, New York: Oxford University Press. 
Bromberger, S. (1966) "Why-Questions," in Mind and Cosmos, ed. R. Colodny, Pittsburgh: 

University of Pittsburgh Press. 
Carnap, R. (1956) "The Methodological Character of Theoretical Concepts," in The Foundations 

of Science and the Concepts of Psychology and Psychoanalysis, ed. H. Feigl and M. Scriven, 

Minneapolis: University of Minnesota Press, pp. 38-76. 
Cartwright, N. (1983) How the Laws of Physics Lie, Oxford: Oxford University Press. 

(1999) The Dappled World, Cambridge: Cambridge University Press. 

Coffa, J. A. (1974) "Hempel's Ambiguity," Synthese 28, pp. 141-63. 

Friedman, M. (19 74) "Explanation and Scientific Understanding," Journal of Philosophy 71, 

pp. 5-19. 
Galison, P. (1987) How Experiments End, Chicago: University of Chicago Press. 
Glymour, C. (1980) Theory andEvidence, Princeton, NJ: Princeton University Press. 
Goodman, N. (1949) "The Logical Simplicity of Predicates," Journal of Symbolic Logic 14. 

(1955) Fact, Fiction, and Forecast, Indianapolis: Bobbs-Merrill. 

Hacking, I. (1983) Representing and Intervening , Cambridge: Cambridge University Press. 
Humphreys, P. (1989) The Chances of Explanation, Princeton, NJ: Princeton University Press. 
Kitcher, P. (1981) "Explanatory Unification," Philosophy of Science 48, pp. 507-31. 
Kripke, S. (19 71) "Naming and Necessity," in Semantics of NaturalLanguages, ed. D. Davidson and 

G. Harman, Dordrecht: Reidel. 
Kuhn, T S. (19 70) The Structure of Scientific Revolutions, Chicago: University of Chicago Press. 

(First published 1962.) 
Nagel, E. (1962) The Structure of Science, New York: Harcourt Brace. 
Pearson, K. (1911) The Grammar of Science, 3rd edn., London: A. & C. Black. 
Putnam, H. (19 73) "Meaning and Reference," Journal of Philosophy 70, pp. 699-711. 
Quine, W. V. (1953) From a Logical Point of View, Cambridge, MA: Harvard University Press. 
Reichenbach, H. (1938) Experience and Prediction, Chicago: University of Chicago Press. 
Salmon, W. (1984) Scientific Explanation and the Causal Structure of the World, Princeton, NJ: 

Princeton University Press. 

(1998) Causality and Explanation, New York: Oxford University Press. 

Suppe, F. (ed.) (19 70) The Structure of Scientific Theories, Urbana, IL: University of Illinois Press. 
Suppes, P. (1967) "What is a Scientific Theory?," in Philosophy of Science Today, ed. A. Danto and 

S. Morgenbesser, New York: Basic Books, 
van Fraassen, B. (1980) The Scientific Image, Oxford: Oxford University Press. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

Nelson Goodman (1906-1998) 


Nelson Goodman, distinguished American philosopher, was born on August 7, 1906 
in Somerville, Massachusetts, and died in Needham, Massachusetts on November 25, 
1998. He received the Bachelor of Science degree from Harvard University in 1928 and 
the Ph.D. from Harvard in 1941. From 1929 to 1940 he operated an art gallery in 
Boston; throughout his life, he was a collector of ancient and modern art. From 1942 
to 1945 he served in the United States Army. Thereafter, he taught for one year at Tufts 
University before his appointment to the faculty of the University of Pennsylvania, 
where he served as associate professor from 1946 to 1951, and then as professor from 
1951 to 1964. From 1964 to 1967 he was the Harry Austryn Wolfson Professor of 
Philosophy at Brandeis University. From 1968 to 1977, he was professor of philosophy 
at Harvard University. 

Goodman's contributions to philosophy are wide-ranging, penetrating, and funda- 
mental. The areas in which he worked include epistemology, philosophy of science, 
philosophy of language, analysis of simplicity, theory of symbols, aesthetics, and meta- 
physics. His work is characterized by unusual originality, typically rejecting conven- 
tional approaches in order to reconceive the problems to be addressed and then 
proposing provocative solutions to them. Thus, for example, he recasts the traditional 
problem of induction so as to require codification, rather than justification of inductive 
practice, thereafter dooming the prospects of purely syntactic or semantic approaches 
to such codification and offering a new pragmatic treatment based on "entrenchment." 
To take another example, he reorients aesthetics as a division of epistemology, con- 
cerned primarily with understanding rather than evaluating works of art, a project that 
leads him to formulate a comprehensive new theory of referential functions embrac- 
ing the literary, pictorial, and other arts as well as the sciences. Finally, rejecting both 
physicalism and phenomenalism, both realism and idealism, he emphasizes the diver- 
sity of equally adequate conflicting conceptualizations for any subject matter, thus 
championing what he calls "irrealism"; the doctrine that there are many worlds if any 
and that worlds are made, not found. 

Goodman's treatments are analytically subtle as well as inventive. His writing is terse 
and telling, making brilliant use both of logic and of metaphor. To summarize his work 
in this brief space is clearly impossible. Instead, we shall give a selective account of his 
main contributions in basic areas of his thought. 



Likeness of meaning 

Goodman's discussions with W. V Quine and Morton White in the late 1940s led to a 
widespread discrediting of the analytic/synthetic distinction. Goodman's paper, "On 
Likeness of Meaning" reflects some of his contribution to these discussions. "Under 
what conditions," he asks, "do two names or predicates in an ordinary language have 
the same meaning?" He considers and rejects such answers to the question as that they 
stand for the same essence or the same image or idea, or that nothing can be conceived 
that satisfies the one but not the other, or that it is impossible that something satisfies 
the one but not the other. Eschewing all reference to essences, images, ideas, conceiv- 
ability, and possibility, he asks whether two predicates have the same meaning if and 
only if they are coextensive, that is, apply to exactly the same things. The answer is no, 
since there are clear cases where words with the same extension (or equally lacking in 
it) differ in meaning, for example "centaur" and "unicorn." Extensional identity is 
indeed a necessary, but not a sufficient condition for sameness of meaning. 

Here, Goodman proposes that it is not only the extensions of the original two words 
themselves that we need to consider (so-called primary extensions) but also the exten- 
sions of their parallel compounds (so-called secondary extensions). A pair of parallel 
compounds is formed by making an identical addition to each of the two words under 
consideration; thus, adding the word "picture" to "centaur" and to "unicorn," we have 
the parallel pair "centaur-picture" and "unicorn-picture." Now, although there are 
neither centaurs nor unicorns, there certainly are centaur-pictures and unicorn- 
pictures and, moreover, they are different. Although the original words have the same 
extension, the parallel compounds differ in extension. Goodman's idea is, then, that the 
difference in meaning between two words is a matter either of their own difference in 
extension or of that of any of their parallel compounds. In general, terms have the same 
meaning if and only if they have the same primary and secondary extensions. 

This idea is generalized to cover cases in which the addition of "picture" yields a com- 
pound with null extension. For example, "acrid-odor-picture" and "pungent-odor- 
picture" have the same (null) extension, neither applying to anything. Compounds can, 
however, be formed by other additions, and Goodman suggests that "description" con- 
stitutes a suffix capable of yielding all the wanted distinctions for every pair of words 
P and 0. For any inscription of the form "a P that is not a 0" is a thing denoted by the 
compound "P-description" but not by the parallel "Q-description." And any inscription 
of "a that is not a P" belongs to the extension of "Q-description" but not to that of 
"P-description." Thus, "pungent-odor-description" and "acrid-odor-description" differ 
extensionally since the first, but not the second, applies to any inscription of the form 
"a pungent odor that is not an acrid odor," and vice versa. Thus, even if all pungent 
odors are acrid and acrid odors pungent, the terms "pungent odor" and "acrid odor" 
differ in meaning. It follows from this proposal, in fact, that no two different words have 
the same meaning. 

To the objection that the compounds used in deriving this radical conclusion are 
"trivial," Goodman replies that when a single form of compound indeed has a different 
extension for every term, "the fact that it has different extensions for two given terms 
is of no striking or special interest. Let us, then, simply exclude every compound for 
which the corresponding compounds of every two terms have differing extensions . . . 



instead of saying that every two terms differ in meaning but that some may not differ 
in interesting ways, we say that two terms differ in meaning only if they differ in certain 
interesting or peculiar ways" (1972: 237-8). 

The new riddle of induction 

The starting point for modern discussions of induction is Hume's denial of necessary 
connections of matters of fact. Effects cannot be simply deduced from their causes, nor 
can predictions be logically demonstrated on the basis of available evidence garnered 
from past experience. What then can be the rational justification of the predictions 
upon which we base all our actions? Hume answers that while there is indeed no deduc- 
tive justification, there is a mental habit which underlies the expectation that phe- 
nomena uniformly conjoined in our past experience will be so conjoined in the future. 
In effect, he offers the uniform past conjunction of events as a mark of those induc- 
tions we find compelling in making our predictions. This idea is also represented by a 
modern version, which has found wide favor among scientists as well as philosophers. 
According to this version, predictions are made in conformity with generalizations that 
have regularly worked in the past. Such congruence with past experience is of course 
no guarantee of future success, but it seems to single out those predictions we adopt at 
any given time. Lacking such guarantee, however, what justification can there be for 
adopting these predictions? 

Goodman argues that the justification of induction, like that of deduction, is only a 
matter of codifying our particular sanctioned inferences, and coordinating them with 
the governing rules of our practice, thus bringing them into agreement with one 
another. "A rule is amended if it yields an inference we are unwilling to accept; an infer- 
ence is rejected if it violates a rule we are unwilling to amend. The process of justifica- 
tion is the delicate one of making mutual adjustments between rules and accepted 
inferences; and in the agreement achieved lies the only justification needed for either" 
(1983: 64). 

How persuasive, then, in codifying our practice is the prevalent regularity doctrine, 
the view that our sanctioned predictions are those in conformity with generalizations 
that have regularly worked so far? Here Goodman introduces the notorious case of the 
green emeralds: suppose all emeralds examined before the present time t are green. We 
predict the next emerald to be examined will be green, since such prediction conforms 
to the generalization that all emeralds are green, a generalization uniformly confirmed 
by all our past evidence. Consider now the predicate "grue," applicable to everything 
examined before t if and only if green but to everything else if and only if blue. Then, 
all emeralds examined before t are not only green, but also grue. Hence, the general- 
ization that all emeralds whatever are grue is supported by no less evidence than the 
generalization that they are all green. The prediction that the next emerald to be exam- 
ined will be blue is thus, by the regularity theory, as confirmed as the prediction that it 
will be green. This theory, whether in Hume or in modern scientific dress, thus fails 
utterly to separate the properly confirmed "green" prediction from the bogus "grue" 
one. "Regularity in greenness confirms the prediction of further cases; regularity in 
grueness does not. To say that valid predictions are those based on past regularities, 
without being able to say which regularities, is thus quite pointless. Regularities are 



where you find them, and you can find them anywhere" (1983: 82). If the old problem 
of justifying our inductive practice has indeed been supplanted by "the new riddle of 
induction" asking for a principled distinction between valid and invalid predictions (or 
more generally, "projections"), we still have a long way to go. 

Goodman's solution depends on utilizing knowledge typically not used in attempts 
to interpret induction. In particular, he presumes some knowledge of past projections, 
that is, of hypotheses that have been actually projected in the past, "adopted after some 
of [their] instances have been examined and determined to be true, and before the rest 
have been examined" (1983: 87). Now, when we consult the record of past projections, 
we find that "green" has clearly been projected much more often than "grue"; it is much 
better entrenched than the latter. The entrenchment of a predicate flows from the actual 
past projections of it and of all other coextensive predicates. While in a sense it is thus 
the class that is entrenched, it becomes so only through the projection of terms that 
determine it. 

Goodman, upon this basis, elaborates a subtle general theory of projection which 
would take us too far afield to characterize here. But the main point to be noted is that 
he hopes to have gone beyond the regularity theory by appealing to regularities in our 
linguistic habits. "Like Hume, we are appealing here to past recurrences, but to recur- 
rences in the explicit use of terms as well as to recurrent features of what is observed. 
Somewhat like Kant, we are saying that inductive validity depends not only upon what 
is presented but also upon how it is organized; but the organization we point to is 
effected by the use of language and is not attributed to anything inevitable or 
immutable in the nature of human cognition. To speak very loosely, I might say that in 
answer to the question what distinguishes those recurrent features of experience that 
underlie valid projections from those that do not, I am suggesting that the former are 
those features for which we have adopted predicates that we have habitually projected" 
(1983: 97). 

Constructionalism and nominalism 

In his approach to philosophy, Goodman was greatly influenced by the constructional- 
ism exhibited in Rudolf Carnap's Der logische Aujbau der Welt, whereby Carnap strove to 
reduce all concepts to the immediately given, with the help of the new mathematical 
logic (see CARNAP). Indeed, in his monumental The Structure of Appearance, Goodman 
includes a detailed exposition and critique of Carnap's system before developing 
an alternative phenomenalistic system taking qualia as atoms rather than Carnap's 

In addition to offering acute analyses of qualia and their concretion, of size, shape, 
order, measure, and time, Goodman devotes the first part of his book to the theory of 
constructional systems. Here he argues that not only is intensional identity too strong 
as a requirement for constructional definition; but even extensional identity is too 
strong, since it precludes alternative adequate systematizations, e.g. taking points as 
certain classes of volumes or taking points as certain pairs of lines. He proposes instead 
a criterion of "extensional isomorphism," which he explains as allowing a given term 
to be definable alternatively "by any of several others that are not extensionally iden- 
tical with one another" (1977: 17). The criterion, which must be applied to the whole 



set of definitions of the system, nevertheless provides for the truth-value preserving 
character of the translations of all sentences we care about. 

In his constructionalism, Goodman is concerned both with system and with sim- 
plicity. Indeed, these two concerns are, as he argues, identical. "The purpose of con- 
structing a system is to interrelate its predicates. The same purpose is served by 
reducing to a minimum the basis required. Every definition at once both increases the 
coherence of the system and diminishes the number of predicates that need to be taken 
as primitive. Thus the motive for seeking economy is not mere concern for superficial 
neatness. To economize and to systematize are the same" (1977: 48). 

Goodman's attitude toward constructional systems issues in a passionate defense of 
the Aufbau against the charge that it is abstract, static, and bloodless, a mere caricature 
of experience. On the contrary, such a charge is in effect an attack against philosophy in 
general, for all philosophy involves conceptualization, abstraction, and systematization, 
the effort to map experience, not to duplicate it. 'A map is schematic, selective, conven- 
tional, condensed, and uniform . . . The map not only summarizes, clarifies, and sys- 
tematizes, it often discloses facts we could hardly learn immediately from our 
explorations." Goodman emphasizes the fact that different maps are useful for different 
purposes. "Let no one suppose that if a map made according to one scheme of projec- 
tion is accurate then maps made according to alternative schemes are wrong" (1972: 
15-16). This point is worth remarking since Goodman's phenomenalistic basis in 
The Structure of Appearance is not, as has often been mistakenly supposed, a matter of 
philosophical conviction precluding other, for example, physicalistic bases; he insists 
that while a constructional system may be adequate and illuminating, it can by no 
means claim a monopoly of wisdom. And he sees no virtue in claims of epistemological 
priority made on behalf of either phenomenalistic or physicalistic systems. 

In constructing systems Goodman insists that the logic employed is not mere neutral 
machinery. Interpreted use of the calculus of classes - employing variables taking 
classes as values - commits the system not only to the individuals explicitly acknowl- 
edged, but also to classes of classes, etc. of these, without limit. "The nominalistically 
minded philosopher like myself," he declares, "will not willingly use apparatus that 
peoples his world with a host of ethereal, platonic, pseudo-entities. As a result, he will 
so far as he can avoid all use of the calculus of classes, and every other reference to 
non-individuals, in constructing a system" (1977: 26). Goodman's principle is "enti- 
ties differ only if their content at least partially differs." He offers the following example: 
'A class (e.g. that of the counties of Utah) is different neither from the single individual 
(the whole state of Utah) that exactly contains its members nor from any other 
class (e.g. that of acres of Utah) whose members exactly exhaust this same whole" 

A pioneering paper by Goodman and W. V. Quine is "Steps Toward a Constructive 
Nominalism" (1947) in which the authors, starting from a renunciation of abstract 
entities, proceed to offer ingenious nominalistic translations of a variety of non- 
nominalistic statements (see QUINE). They soon, however, concede their inability thus 
to translate all of mathematics and they therefore suggest a different way of saving 
mathematics nominalistically: to devise a nominalistic syntax capable of describing the 
syntactic rules by which mathematical inscriptions, as concrete uninterpreted marks, 
are manipulated. 



In this enterprise, they succeed in defining a suitable nominalistic syntax language 
for describing the object language of mathematics, with its ingredient notions of 
"axiom," "rule," "proof," and "theorem." In conclusion, they declare "Our position is 
that the formulas of platonistic mathematics are, like the beads of an abacus, conve- 
nient computational aids which need involve no question of truth. What is meaning- 
ful and true in the case of platonistic mathematics as in the case of the abacus is not 
the apparatus itself, but only the description of it: the rules by which it is constructed 
and run. These rules we do understand, in the strict sense that we can express them in 
purely nominalistic language" (1972: 198). 

Further discussion of nominalism, including the later differing interpretations 
of Goodman and Quine, as well as elaborations and answers to various objections, can 
be found in Goodman's paper, "A World of Individuals" (1972: 155-72). Here, he 
declares that his nominalism is specifically "the refusal to recognize classes" (p. 156), 
requiring that all entities admitted to a system, no matter what they are, be treated as 

Theory of symbols 

In Languages of Art, Goodman develops a theory of symbols that encompasses but 
ranges far beyond what is typically considered aesthetics, while giving only minor 
attention to artistic value and criticism. His primary concern is to develop a systematic 
approach to the functioning of symbols not only in the arts but also in the sciences and 
in ordinary contexts as well, for example, in the everyday use of labels and samples. 
And in the realm of the arts proper, he discusses music as well as painting, literary arts 
as well as dance and architecture. His key emphasis is on cognitive function and to this 
end he devises a general theory which he employs in characterizing notations, sketches, 
scripts, and paintings. In related discussions, his book offers an account of forgery of 
works of art, a theory of metaphor and a categorization of figures of speech, a treat- 
ment of fictional expressions, and a comprehensive interpretation of reference as 
including not only denotation, but also exemplification and expression. 

A swatch of cloth in a tailor's shop is typically used to exemplify certain of its prop- 
erties, i.e. its color and texture, but not its size or shape. To serve thus, it must possess 
these properties, but also refer to them. Exemplification is a symbolic function by which 
the sample stands for a property it possesses. Expression implies metaphorical exempli- 
fication: if a picture expresses sadness, it is itself metaphorically sad and also refers to 
sadness. With these two notions at hand, Goodman is able to treat works of art not 
merely as objects of reference but - even where nonrepresentational - as referring 
symbols in their own right. And the ingredient idea of metaphor vastly expands the 
resources available for treating the wide expressive ranges of works of art. 

Goodman's treatment of metaphor has been widely noted. "Where there is 
metaphor," he writes, "there is conflict: the picture is sad rather than gay even though 
it is insentient and hence neither sad nor gay. Application of a term is metaphorical 
only if to some extent contra-indicated" (1976: 69). Along with such contra- 
indication, there must also be attraction, or aptness. The metaphorical use of "sad" 
implies that there are two ranges for the term "sad," but that these two ranges do not 
simply comprise an ambiguity. In mere ambiguity, the separate uses of the term are 



independent. "In metaphor, on the other hand, a term with an extension established 
by habit is applied elsewhere under the influence of that habit; there is both departure 
from and deference to precedent. When one use of a term precedes and informs another, 
the second is the metaphorical one" (p. 71). 

The key technical discussion in the book presents a theory of notation, for example, 
a musical score. The main function of such a notation, says Goodman, is to provide 
authoritative identification of a work from performance to performance, whatever 
other uses it may have. "What is required is that all and only performances that comply 
with the score be performances of the work" (1976: 128). With this principle as a 
guide, Goodman develops five requirements for a notational system: unambiguity and 
syntactic and semantic disjointness and differentiation. 

To give a precise account of these requirements would exceed the limits of this space, 
but a brief summation of their import is stated by Goodman as follows: "A system is 
notational, then, if and only if all objects complying with inscriptions of a given char- 
acter belong to the same compliance class and we can, theoretically, determine that 
each mark belongs to, and each object complies with inscriptions of, at most one par- 
ticular character" (1976: 156). Using this notion of notation, Goodman is able then to 
characterize other things than, for example, scores. A script, for example "is a charac- 
ter in a notational scheme and in a language but, unlike a score, is not in a notational 
system. The syntactic but not all the semantic requirements are met ... [it is] a 
character in a language that is either ambiguous or lacks semantic disjointness or 
differentiation" (pp. 199, 201). 

Rather than offering a definition of the aesthetic, Goodman suggests four symptoms 
that tend to be present in aesthetic experience: syntactic density, semantic density, syn- 
tactic repleteness, and exemplificational character. The first two call for maximum sen- 
sitivity of discrimination, the third calls for such effort along many dimensions, while 
the fourth is shown by concern with properties exemplified by a symbol, not merely 
with things the symbol denotes. Largely as a result of his analysis, Goodman empha- 
sizes the affinities between art and science, despite their differences. The difference 
between them is not that between feeling and fact or truth and beauty "but rather a 
difference in domination of certain specific characteristics of symbols" (1976: 264). 


In Goodman's Ways of Worldmaking, he speaks of versions as including depictions as 
well as descriptions, works of art as well as works of science. And he insists that there 
are conflicting right world-versions, rather than a single world underlying the Tight- 
ness of all. He considers his view on these matters as 

belonging in that mainstream of modern philosophy that began when Kant exchanged the 
structure of the world for the structure of the mind, continued when C. I. Lewis exchanged 
the structure of the mind for the structure of concepts, and that now proceeds to exchange 
the structure of concepts for the structure of the several symbol systems of the sciences, 
philosophy, the arts, perception, and everyday discourse. The movement is from unique 
truth and a world fixed and found to a diversity of right and even conflicting versions or 
worlds in the making. 



The view that emerges, he says, can perhaps be described as "a radical relativism under 
rigorous restraints, that eventuates in something akin to irrealism" (1978: x). 

Goodman takes his adversary to be "the monopolistic materialist or physicalist who 
maintains that one system, physics, is preeminent and all-inclusive, such that every 
other version must eventually be reduced to it or rejected as false or meaningless" 
(1978: 4). Were all versions reducible to a single one, that one might be plausibly con- 
sidered the only truth about the one world. But such reducibility is a chimera; the claim 
of reducibility to physics is "nebulous since physics itself is fragmentary and unstable 
and the kind and consequences of reduction envisaged are vague . . . The pluralisms 
acceptance of versions other than physics implies no relaxation of rigor but a recogni- 
tion that standards different from yet no less exacting than those applied in science 
are appropriate for appraising what is conveyed in perceptual or pictorial or literary 
versions" (p. 5). 

Doesn't the Tightness of these various versions imply an underlying world that 
makes versions right? No, says Goodman, 

we might better say that "the world" depends on Tightness. We cannot test a version by 
comparing it with a world undescribed ... all we learn about the world is contained in 
right versions of it; and while the underlying world, bereft of these, need not be denied to 
those who love it, it is perhaps on the whole a world well lost . . . For many purposes . . . 
just versions can be treated as our worlds. (19 78: 4) 

Worlds are made through the making of versions, but versions cannot be made 
any way we like. Goodman denies that his acceptance of many right world-versions 
implies "that anything goes, that tall stories are as good as short ones, that truths are 
no longer distinguished from falsehoods." Although it is true, he says, that "we make 
worlds by making versions," we cannot do so at random or by whim (1978: 94). "Of 
course," he says, 

we want to distinguish between versions that do and those that do not refer, and to talk 
about the things and worlds, if any, referred to; but these things and worlds and even the 
stuff they are made of - matter, anti-matter, mind, energy, or whatnot - are fashioned 
along with the things and worlds themselves, (p. 94) 

The making of worlds is brought about by the making of versions and the multiple 
worlds thus made "are just the actual worlds . . . answering to true or right versions. 
Worlds possible or impossible supposedly answering to false versions," says Goodman, 
"have no place in my philosophy" (1978: 94). 

Goodman's relativism on the issue of worldmaking is familiar from his earlier work 
on constructionalism, where, as we have seen, he insists on honoring the conflicting 
definitions that can be offered for the very same terms. The criterion of extensional iso- 
morphism, as we have noted above, is a more relaxed standard than either synonymy 
or identity, but it nonetheless imposes clear and definite restraints. 

Similarly, his theory of worldmaking, paradoxical as it may seem, recognizes a vast 
array of conflicting versions, rather than a single underlying world, yet insists that 
there are clear distinctions to be drawn between right and wrong versions. Goodman's 



relativism must therefore be sharply distinguished from nihilism, subjectivism, and 
cultural relativism. His hospitality to variant and opposed conceptualizations is allied 
with a dedication to the highest standards of logical rigor. 


Works by Goodman 

1947 (with Quine, W. V.): "Steps Toward a Constructive Nominalism," Journal of Symbolic Logic 
12, pp. 105-22. (Reprinted in Goodman 1972, pp. 173-93.) 

1949: "On Likeness of Meaning," Analysis 10, pp. 1-7. (Reprinted in Goodman 1972, 
pp. 221-30.) 

1953: "On Some Differences about Meaning," Analysis 13, pp. 90-6. (Reprinted in Goodman 
1972, pp. 231-8.) 

1963: "The Significance of Tier logische Aufbau der Welt," in The Philosophy of Rudolf Carnap, ed. 
P. A. Schilpp, La Salle, IL: Open Court and London: Cambridge University Press. (Reprinted in 
Goodman 1972, pp. 5-23.) 

1972: Problems and Projects, Indianapolis: Bobbs-Merrill. 

1976: Languages of Art, 2nd edn., Indianapolis: Hackett. (First published Indianapolis: Bobbs- 
Merrill, 1968.) 

1977: The Structure of Appearance, 3rd edn., Dordrecht: Reidel. (First published Cambridge, MA: 
Harvard University Press, 1951.) 

1978: Ways of Worldmaking, Indianapolis: Hackett. 

1983: Fact, Fiction and Forecast, 4th edn., Cambridge, MA: Harvard University Press. (First pub- 
lished 1955 by Harvard University Press.) 

1984: Of Mind and Other Matters, Cambridge, MA: Harvard University Press. 

1988 (with Elgin, Catherine Z.): Reconceptions in Philosophy and Other Arts and Sciences, 
Indianapolis: Hackett. 

Works by other authors 

Elgin, C. Z. (1983) With Reference to Reference, Indianapolis: Hackett. 

(ed.) (1997) The Philosophy of Nelson Goodman, 4 vols, New York and London: Garland 

Publishing. (Contains articles by various authors on Goodman's constructionalism, theory of 
induction, philosophy of art, and theory of symbols.) 

Howard, V. A. (1982) Artistry, Indianapolis: Hackett. 

McCormick, P. J. (ed.) (1996) Starmaking, Cambridge, MA: MIT Press. (Contains critical discus- 
sions of Goodman's irrealism, by Hempel, Putnam, Scheffler, and Goodman.) 

Scheffler, I. (1963) The Anatomy of Inquiry, New York: Alfred A. Knopf. 

(199 7) Symbolic Worlds, Art, Science, Language, Ritual, Cambridge: Cambridge University 


Stalker, D. (1994) Grue: The New Riddle of Induction, Chicago: Open Court. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

H. L.A.Hart (1907-1992) 


Herbert Lionel Adolphus Hart was born of Jewish parents in Yorkshire, England and 
was educated at New College, Oxford. After graduating with a First in Greats, Hart was 
called to the Bar as a Chancery barrister in London. He spent the next eight years build- 
ing a successful legal practice, specializing first in property conveyancing, trust draft- 
ing, and tax planning, and then moving on to court work and advising. Although his 
interests quickly turned from law to philosophy, Hart continued to practice and, in fact, 
during this period he declined an invitation to teach philosophy in Oxford. His legal 
career, however, was cut short by World War II. While working with British Intelligence, 
Hart met Gilbert Ryle and Stuart Hampshire, from whom he learned of the new trends 
in philosophy. When the war ended, Hart left his law practice and returned to Oxford. 
In 1952, Hart was elected to the Chair of Jurisprudence, a somewhat surprising 
appointment given that he did not have a degree in either law or philosophy and had 
published little by that point. He occupied that chair, however, with great distinction, 
publishing several seminal works in legal theory, including his masterpiece, The Concept 
of Law, in 1961. 

Hart is perhaps best known for his vigorous and sophisticated defense of the 
doctrine known as legal positivism. In its broadest sense, legal positivism is a 
theory about the nature of law that denies any necessary connection between 
legality and morality. No stipulation is made that, in order to count as law, a norm 
must possess any moral attributes. Legal positivists, therefore, believe that it is possible 
for a legal system to recognize a rule as legally valid even if it happens to be unjust. 
This analytic separation between the legal and the moral was captured by John 
Austin when he said: "The existence of law is one thing; its merit or demerit is another" 
(1954: 184-5). 

In an effort to cleanse analytic jurisprudence of its moral content, every legal posi- 
tivist before Hart thought it necessary to recast the basic legal concepts of obligation, 
rule, validity, and authority in terms of sanctions. Austin, for example, believed that legal 
rules are nothing more than orders backed by threats of sanctions issued by the sover- 
eign. Sovereignty, in turn, was understood in terms of coercive power, the sovereign 
being the one in a group who has the power to elicit habitual obedience from every one 
and who habitually obeys no one. Someone is under a legal obligation to act, on this 
view, if they are likely to be sanctioned for failing to act. 



Hart was firmly committed to the analytic separation of law and morality, but 
thought that these sanction-centered theories distorted and concealed the various ways 
in which the law guides conduct. For example, Hart pointed out that there are many 
legal rules that lack sanctions, in the sense that no penalties are imposed as a result of 
non-conformity with them. If a person drafts a will but fails to have it signed by two 
witnesses, that person is not sanctioned for the inadequate attention paid to the testa- 
mentary rules. He has simply failed to form a valid will and his actions lack legal effect. 

Sanction-centered theories fall short, according to Hart, because they treat all legal 
rules as if their sole function is to discourage undesirable behavior. Their paradigm is 
the criminal law, where the rules impose duties to act or forbear from certain behavior 
and specify sanctions in the event of disobedience. However, not only do the rules 
related to valid will or contract formation lack sanctions, but, as Hart observed, it does 
not even make sense to speak of obeying or disobeying them. These rules do not impose 
duties; they instead confer powers. Their function is not to discourage people from acting 
in ways that they otherwise might wish, but to give them facilities for realizing their 

The effacement of power-conferring rules is especially problematic with respect to 
those rules that confer legal powers on public officials. Without such rules, Hart noted, 
sanction-centered theories cannot account for the self-regulating nature of legal insti- 
tutions: it is a defining feature of law, as opposed to pre-legal social systems, that its offi- 
cials are empowered to change the rules and to resolve the disputes that may arise 
under them. 

Hart also believed that these theories give a misleading picture of the nature of the 
law's normativity. On the sanction-centered approach, the only reasons for action that 
the law provides are threats of sanctions. This ignores what Hart called the internal point 
of view, which is the perspective of those who treat the rules as standards of acceptable 
conduct. In every legal system, Hart claimed, some members of the group treat the 
rules not just as threats, or predictions of what courts will do, but as guides to their 
conduct and standards for the evaluation of others - as norms that obligate and 
empower, not merely oblige. 

By emphasizing the internal point of view, Hart was not simply criticizing fellow 
legal positivists for neglecting an obvious fact, i.e. that at least some people in some cir- 
cumstances are motivated by the law qua law, instead of sanctions. Rather, Hart was 
also mounting a methodological offensive against the crude scientism of some of his 
contemporaries. For example, Alf Ross, the Scandinavian Legal Realist, based his legal 
positivism on his commitment to logical positivism and believed that, if jurisprudence 
is to have empirical content, legal concepts must be operationalized in purely behav- 
ioristic terms. By contrast, Hart believed that theories of law must make essential 
reference to the attitudes of legal actors because the law is a social practice. In order to 
analyze the practice, it is not enough to record regularities of behavior; one must 
understand how the participants understand it. Hart's introduction of the internal 
point of view thus inaugurated the hermeneutic turn in jurisprudence, where the law is 
studied from the inside, that is, from the perspective of those who live under, and directly 
experience, the law. 

By engaging in this hermeneutic enterprise, Hart was not, however, giving up on a 
naturalistic approach to legal theory. Indeed, Hart believed that the internal point of 


H. L. A. HART 

view allowed the legal positivist to anchor rules in social facts. According to Hart, a 
social rule in a community exists whenever a sufficient number of people engage in a 
practice from the internal point of view. This internal aspect of rules is manifested exter- 
nally in conforming behavior, as well as criticisms that attend deviations from the prac- 
tice and the use of normative language such as ought, must, and obligation to express 
such disapprobation. The existence of a rule, therefore, is firmly rooted in the natural 
world, that is, in regularities of behavior motivated by the appropriate critical attitude. 

Hart's theory of social rules forms the foundation of his approach to law. According 
to Hart, at the root of every legal system is a social rule of a special sort, which he called 
the rule of recognition. This rule imposes a duty on courts to apply rules that bear certain 
characteristics. In the American system, for example, the rule of recognition requires 
judges to apply the rules duly enacted by Congress. The rule of recognition, therefore, 
sets out the criteria of legal validity, that is, those criteria that a rule must possess in 
order to be law. 

The rule of recognition is what Hart called a secondary rule: it is a rule about other 
rules. It is also an ultimate rule: it exists because it is accepted by judges from the inter- 
nal point of view, not in virtue of its validation by another rule. The primary rules, by 
contrast, owe their existence to the rule of recognition, and not to any guidance that 
they might engender. 

In addition to the rule of recognition, Hart argued that every legal system contains 
two other secondary rules. The rule of change confers the power on legislative bodies to 
modify the primary rules, whereas the rule of adjudication confers the power on courts 
to adjudicate whether the primary rules have been followed or violated. 

By understanding the law as the union of primary and secondary rules, Hart intro- 
duced what might be called a rule-centered theory of the law. On this model, the law 
guides conduct not by issuing naked threats, but by providing rules that impose duties 
and confer powers. The basic legal concepts are also understood in terms of rules, not 
sanctions. A rule is valid in a legal system when the rule bears those characteristics set 
out in that system's rule of recognition. An act is legally obligatory, in turn, when it is 
required by a legally valid rule. A person has supreme legal authority when the sec- 
ondary rules of the system confer legal power on that person and no other has been 
conferred a greater power. Even the concept of a sanction is rendered in terms of rules, 
for a sanction is not simply a cost imposed by the law, but, unlike a tax, is a penalty 
exacted because a rule has been violated. 

Hart did not think that privileging the concept of a rule compromised the 
analytic separation of law and morals. In his model, a primary legal rule exists just in 
case it is validated by that system's rule of recognition. There is no demand that the 
criteria of legal validity set out make reference to the rule's moral properties. It is 
possible, and regrettably often the case, that a legal rule exists even though, from a 
moral point of view, it should not. And while it is true that the rule of recognition would 
not exist unless judges accept it from the internal point of view, this does not mean that 
they judge it morally acceptable or that it is morally acceptable for them to treat it in 
this way. 

Despite Hart's insistence that law be seen as a system of rules, he did not think that 
judges are always guided by these rules when they decide cases. In his view, courts are 
not simply the passive servants of the legislature or of tradition, restricted to applying 



the rules laid down in advance, but are active players in the creation and development 
of the law. Judges do not always find the law; they sometimes make it as well. 

Hart, however, was not disturbed by the fact of judicial legislation. He thought that 
judges should be given free rein to decide some cases, as it enables them to fashion 
sensible solutions to unforeseen problems. Moreover, given the inherent limitation of 
natural languages, he believed that judicial legislation was unavoidable. According 
to Hart, all general terms in natural language (e.g. vehicle) contain a core of settled 
meaning (e.g. car) and a penumbra where the reference class is ill-defined (e.g. tractor). 
When a case falls into the core of a general term of the rule, the rule applies and the 
judge is legally obligated to apply the rule. However, when in the penumbra, the law 
runs out and the judge must exercise his discretion. By necessity, the judge cannot find 
the law, because there is no law to find, and hence must make new law. 

Although sounding sensible enough, Hart's recognition, and sympathetic accept- 
ance, of judicial legislation has been attacked by his chief critic, and successor to the 
Chair in Jurisprudence, Ronald Dworkin. In Dworkin's view, the role of a 
judge is to vindicate the legal rights of the parties and this can only be accomplished if 
the law completely regulates the judge's behavior in every case. Dworkin faulted 
Hart for counting as law only rules that have social pedigrees, such as legislation or 
custom, and ignoring the mass of implicit law represented by moral principles 
that justify the pedigreed rules and that determine the legally correct answer when 
these rules run out. 

By arguing that, in every case, there is a right answer, Dworkin was not only chal- 
lenging Hart's theory of adjudication but also his claim that law and morality were 
conceptually distinct. For if the legally correct answer is determined in part by norms 
whose only claim to legal validity is their moral validity, then it would seem that moral- 
ity would be a determinate of legality, contrary to legal positivistic strictures. 

In the Postscript to the second edition of The Concept of Law, published posthu- 
mously, Hart agreed with Dworkin that judges are often legally obligated to apply moral 
principles that lack pedigrees, and that when judges act on them, they are applying 
existing law. However, Hart believed that such a position was consistent with legal pos- 
itivism, for he saw no reason why the rule of recognition could not validate a norm 
based on its moral properties. Legal positivists, according to Hart, only claim that a rule 
of recognition need not validate a norm on the basis of its moral content, not that it 
cannot. Even when the rule of recognition did validate principles on the basis of their 
moral content, Hart doubted that these principles would indicate a unique result in 
every case, thus leaving ample room for the exercise of judicial discretion. 

In separating law from morals, Hart did not mean to preclude moral criticism of the 
law. Quite the contrary, Hart was a vocal and influential critic of many aspects of the 
criminal law, especially the prohibitions on so-called private vices. In Law, Liberty and 
Morality, Hart attacked the doctrine known as legal moralism, the belief that society has 
the right to use the criminal law to enforce its moral code. Lord Devlin had argued that 
social cohesion is possible only when a common code of morality is respected by all, 
and the flouting of that code, even in private, threatens such cohesion and, in turn, 
society's very existence. Hart noted that Devlin failed to produce any evidence sup- 
porting his causal claims, and doubted whether any could be mustered. More impor- 
tantly, he argued that a society that criminalizes behavior that the majority finds 


H. L. A. HART 

offensive is not a society that respects liberty. To respect liberty, a society must protect 
the right of individuals to choose their own lifestyle, even when it does not approve of 
the lifestyle they end up choosing. The liberty to act only in ways that others like is, as 
Hart pointed out, liberty in name only. 

In contrast to most of his contemporaries, Hart eschewed grand moral theories 
in favor of a more commonsense approach to normative analysis, which borrowed 
elements from both the Utilitarian and Kantian traditions. For example, Hart thought 
that the justifying aim of punishment is the deterrence of crime. Yet, he also 
believed that this pursuit must yield to the demands of justice, so that it is wrong to 
punish people for crimes they did not commit or could not have helped committing. 
He was thus critical of the attempts to increase the efficiency of the criminal law by 
eliminating many of the traditionally recognized excuses, such as the restrictions 
on the use of the insanity defense and the introduction of crimes of strict liability and 

Although Hart recognized that the availability of excuses might allow some to feign 
incapacity or mistake and thus evade responsibility, he nevertheless thought that the 
costs are slight compared to the benefits. Not only is it fundamentally unfair to punish 
those who could not have helped doing what they did, but, as Hart pointed out, a system 
of excuses places individuals in control of their destinies. For when the law only pun- 
ishes people for actions they can avoid, people can avoid being punished. As long as 
individuals never choose to break the rules, the law will let them go about their lives. As 
a result, individuals need not fear that they will unwittingly bring the wrath of the law 
down upon themselves; they can rely on the fact that the law will excuse behavior that 
was not, in some suitable sense, a product of choice. 

It is a mistake, Hart concluded, to think that reducing crime by eliminating excuses 
will lead to an increase in security. When excuses are unacceptable, people are unable 
to predict the consequences of their actions. A world that is unknowable and un- 
controllable is a world in which no one is secure. 


Works by Hart 

1955: 'Are there Any Natural Rights?," Philosophical Review 64, pp. 175-91. (Reprinted in 
Political Philosophy, ed. A. Quinton, Oxford: Oxford University Press, 1967, pp. 53-66.) 

1958: "Positivism and the Separation of Law and Morals," Harvard Law Review 71, pp. 593-629. 

1959 (with Honore, A. M.): Causation in the Law, Oxford: Clarendon Press. 

1961: The Concept of Law, 1st edn., Oxford: Clarendon Press. (The 2nd edn., 1994, includes a 
"Postscript," which is a reply to critics.) 

1963: Law, Liberty and Morality, London: Oxford University Press. 

1968: Punishment and Responsibility, Essays in the Philosophy of Law, Oxford: Clarendon Press. 

1982: Essays on Bentham, Oxford: Clarendon Press. 

1983: Essays in Jurisprudence and Philosophy, Oxford: Clarendon Press. 

Works by other authors 

Austin, J. (1954) The Province of Jurisprudence Determined, London: Weidenfeld and Nicolson, 
pp. 184-5. 



Dworkin, R. (1977) "The Model of Rules I" and "The Model of Rules II," in Taking Rights Seriously, 

Cambridge, MA: Harvard University Press. 
Fuller, L. (1958) "Positivism and Fidelity to Law: A Reply to Professor Hart," Harvard Law Review 

71, p. 630. 
MacCormick, N. (1981) H. L. A. Hart, Stanford: Stanford University Press. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

C. L. Stevenson (1908-1979) 


Stevenson's major contribution to philosophy was his development of emotivism, a 
theory of ethical language according to which moral judgments do not state any sort 
of fact, but rather express the moral emotions of the speaker and attempt to influence 

Stevenson's emotive theory of ethical language 

Stevenson always stressed that his work did not include any substantive moral judg- 
ments, but rather comprised "analytic ethics," or what is now commonly called 
"metaethics," the branch of moral theory that is about ethics and ethical language. 

What do we mean when we say that something is good or bad, or right or wrong? 
On the face of it, we are describing, attributing to the thing some property, goodness or 
badness, or Tightness or wrongness. What could these properties be? How do we find 
out about them? Much of philosophical moral theory explores various answers to these 
questions. Stevenson thought that questions about the nature of moral properties were 
misplaced. Our moral judgments do not, at least primarily, describe at all. Uttering 
moral sentences has a different function: to express emotions, and to influence or invite 
others to share them. All of his main contributions appeared in Ethics and Language, 
1944, and a collection of papers, Facts and Values, 1963. 

Distinguish between expressing a certain state of mind and saying that one is in it. If 
I say, 'Ann Arbor is in Michigan," I express my belief that Ann Arbor is in Michigan, 
but I do not say that I believe such a thing. For what makes what I said true? Not that 
I really do believe that Ann Arbor is in Michigan; only the fact that Ann Arbor really 
is in Michigan. Stevenson's theory of ethical language, in a nutshell, was that when 
I say, "Inequality is bad," I have expressed a certain negative moral attitude toward 
inequality, though I have not said that I have it. It should be clear why Stevenson 
stressed that his theory was "analytic" or metaethical, and did not contain any sub- 
stantive moral judgments. For by claiming that moral judgments serve to express emo- 
tions, he had not expressed his own moral emotions at all. 

Besides expressing the speaker's attitude, Stevenson said, moral statements also 
"create an influence," they invite the audience to share in the emotion expressed. Thus, 
"x is good" is akin to "Let us approve of x." Moral exhortation, after all, is commonly 



used to try to persuade the audience to share the speaker's suggestions, and moral judg- 
ment is often a call to action. Furthermore, in context, ethical statements can come to 
have some secondary descriptive content; in Victorian England, for example, calling a 
woman "virtuous" implied that she was chaste. So a Victorian moralist could manage 
to describe a woman, and not merely to evaluate her (express his emotional attitude 
toward her and invite others to share it), by calling her "virtuous." 

Some advantages of emotivism 

Stevenson's theory was enormously influential in the middle of the twentieth century. 
Taking its cue from Ayer's short remarks on ethics in Language, Truth and Logic, 
Stevenson's theory added sophistication and subtlety (see AYER). 

In "The Emotive Meaning of Ethical Terms," Stevenson sets out some criteria for a 
successful analysis of moral terms, explaining that what he calls traditional "interest 
theories" of ethical terms fail one or more criteria. These interest theories include the 
views of Hobbes, whom Stevenson understood to have defined "good" to mean "desired 
by me," and Hume, whom he interpreted as defining it to mean "desired by my 

In the first place, we must be able sensibly to disagree about whether something is "good." 
This condition rules out Hobbes's definition. For consider the following argument: "This is 
good." "That isn't so; it's not good." As translated by Hobbes, this becomes: "I desire this." 
"That isn't so, for I don't ..." 

In the second place, "goodness" must have, so to speak, a magnetism. A person who rec- 
ognizes X to be "good" must ipso facto acquire a stronger tendency to act in its favour than 
he otherwise would have had. This rules out the Humean type of definition. For accord- 
ing to Hume, to recognize that something is "good" is simply to recognize that the major- 
ity approve of it. Clearly, a man may see that the majority approve of X without having, 
himself, a stronger tendency to favour it . . . 

In the third place, the "goodness" of anything must not be verifiable solely by use of the 
scientific method. "Ethics must not be psychology." This restriction rules out all of the 
traditional interest theories, without exception. 

Emotivism appears to be well prepared to satisfy these criteria. First, people can gen- 
uinely disagree when one states that X is good and the other states that X is not good; 
they are disagreeing in attitude, as Stevenson puts it, not in factual belief, but this is 
genuine disagreement just as plainly as we may disagree when I suggest that we go to 
the movies tonight and you suggest that we go have a few drinks instead. Second, and 
perhaps most significantly, if the judgment that X is good is an expression of favorable 
attitude toward X, then it is clear why anyone making such a judgment will have a ten- 
dency to act in favor of X. Finally, while adding to our knowledge by scientific investi- 
gation may sometimes resolve certain ethical issues, there can be deeper disagreements 
that are left untouched by scientific methods. Our emotional attitudes may differ in the 
face of converging empirical knowledge. 

A further attraction of emotivism is that it dissolves knotty-looking metaphysical 
problems of metaethics. Consider the question of whether moral properties are natural 
properties or some other special sort. G. E. Moore famously argued that moral proper- 



ties could not be natural properties, and later John Mackie argued for skepticism about 
the existence of moral properties on the grounds that they could not be natural ones, 
and the metaphysics and epistemology of non-natural properties is too spooky (or 
"queer," as Mackie said) for sober philosophy. Some metaethicists have tried to show 
how moral properties could be part of the natural world after all, but it is difficult to 
explain just how our linguistic habits and practices could determine just which natural 
property moral wrongness could be, given the wide diversity and disagreement in moral 
values among different people and cultures at different times. Emotivism resolves the 
issue by denying that moral predicates, like "wrong" and "good," serve to pick out prop- 
erties at all. They serve as markers of mood or emotion instead. So the metaphysics of 
alleged moral properties is avoided if we adopt Stevenson's view. 

Some difficulties for emotivism 

Emotivism is not without its difficulties, and the main ones were leveled at Stevenson 
soon after he began to publish his views. One criticism was offered by Brand Blanshard 
in a paper called "The New Subjectivism in Ethics." Blanshard complained that emo- 
tivism has an obviously false implication. When I see a rabbit with its foot caught in a 
trap, f might say (or think) "That's a bad thing." I would then, plausibly, be expressing 
my negative emotion toward the pain of the rabbit. But suppose I then contemplate the 
situation in which I myself become very jaded and cease to care about the suffering of 
sentient animals. Do I (now, actually) say, "Well, in that case, the suffering of the rabbit 
would not be a bad thing at all"? No, of course not. But emotivism implies that this is 
how I should think. So emotivism is false. 

This criticism is instructive, though it is not correct. It illustrates two important 
points about Stevenson's theory. First, the fact that we would ordinarily say one thing 
or another is very important, according to Stevenson's approach. He would never have 
replied, "We might not say such a thing in that case, we might steadfastly deny it, but 
we would be mistaken." His theory was supposed to account for our ordinary judg- 
ments, and not to reform those judgments. So it is important whether Blanshard's 
example really does show that emotivism sometimes contradicts our ordinary ethical 

However, the criticism is unsound, because Stevenson's approach does not, in fact, 
imply that we do or should judge that the suffering of the rabbit would not be at all bad 
if we were jaded and uncaring. To think that it does imply such a thing is to mistake 
emotivism for a poor relation, subjectivism. The subjectivist thinks that "bad" means 
(something like) "apt to cause a negative emotion in me." So to call something bad, 
according to subjectivism, is to say that it causes a negative emotion in oneself. But 
Stevenson took great pains to distinguish his own view from subjectivism, and he gave 
very similar examples to show why subjectivism is incorrect. According to emotivism, 
remember, calling something bad is not saying that it does or doesn't do anything - that 
would be to describe the thing. Ethical language does not (primarily) describe a thing 
or an emotion or the speaker, it expresses the emotion of the speaker. When I contem- 
plate the situation in which I heartlessly feel no sorrow over the rabbit's suffering, I 
(right now, actually) feel rather bad about that, and if I were to express my emotion I 
would say, "That would be a bad thing." 



Probably the most influential criticism of Stevenson, the criticism that later emo- 
tivists (and fellow non-descriptivists, see below) have been most concerned to address, 
was a problem noticed by Peter Geach and John Searle. ft is sometimes called the 
"embedding problem." To put it succinctly, the problem is that even if emotivism really 
does tell us what somebody does when she asserts a simple moral sentence like "ft is 
wrong to kick cats," it does not seem to tell us what such a sentence means. For there 
is more to the meaning of a sentence than the facts about what is accomplished or 
expressed by an assertion of it, since we can use sentences without asserting them, in 
unasserted or embedded contexts. There are many kinds of unasserted contexts. Here are 
a few examples; notice that in no case would someone sincerely uttering the entire 
sentence be asserting that it is wrong to kick cats. 

If it is wrong to kick cats, then it is wrong to kick Tibbies. 

Either it is wrong to kick cats, or there is nothing wrong with kicking people. 

f wonder whether it is wrong to kick cats. 

Do you mean to say that it is wrong to kick cats? 

Many other kinds of examples could be given, but the idea is clear enough. Critics 
of emotivism point out that what Stevenson said about the emotive meaning of ethical 
terms does not seem to explain how a sentence like "ft is wrong to kick cats" embeds 
into these complex contexts. What, that is to say, does the sentence contribute to the 
complex whole? One thing is clear enough: someone uttering any of the four example 
sentences above could not be said to be expressing a negative emotion toward kicking 
cats. So something more must be said. Stevenson himself never seems to have taken 
this problem to heart, so he never said much of anything by way of reply. But some 
later non-descriptivists have said more (see below). 

The embedding problem may appear to be a kind of technicality, and perhaps 
it is, though many philosophers have taken it very seriously. The final criticism f will 
mention seems to cut deeper into the spirit of emotivism. Stevenson said, and emotivism 
gains much of its plausibility from this idea, that a person who sincerely asserts or 
believes a moral judgment must necessarily feel some sort of emotional tug, so that 
whoever judges something good must be emotively in favor it, and whoever 
judges something bad must be against it or inclined to avoid, or would try to eliminate 
it. But we may wonder whether this claim is true, isn't it possible to judge sincerely 
that something is good, but feel no sympathy or other "pro-attitude" toward it 
whatsoever? There is no uncontroversial answer. Some find it obvious that such a 
thing is possible, while others are at least at first inclined to wonder what the questioner 
could possibly have in mind. But if we tell a background story it starts to seem very 
plausible that Stevenson may have overstated the connection between moral judgment 
and emotion. 

Surely it is imaginable that someone could be a self-avowed and sincere amoralist. 
Such a person would have no interest at all in moral values or rules, and might 
even be perfectly forthright in admitting so. Yet amoralists could surely learn to recog- 
nize which things are good and bad, even if the normal concern with such things might 
seem quaint or misguided to them. So they could with perfect sincerity and under- 
standing manage to judge that giving to charity is morally good, or that breaking 
promises is bad, and they could make those judgments without any emotion or 



motivation or tendency to promote the "good" things or discourage the "bad" ones. All 
of this seems possible. Doesn't it show that emotivism is a mistaken theory? 

Perhaps not. In the paragraph above, the words "good" and "bad" are in quotation 
marks. It is plausible that amoralists use these and other moral words in what R. M. 
Hare called the "inverted commas sense," really mentioning them rather than using 
them. Amoralists cannot say (sincerely, at least) that charity is good, so instead they 
say that charity is what most folk call "good." We (moralists or amoralists) can certainly 
mention emotive words without expressing their emotive meanings. 

Some related theories 

Hare's theory is not emotivist, though it is a close ally. According to Hare, the main 
function served by moral judgments is prescription (see HARE). So Hare agrees with 
Stevenson that we do not fundamentally describe things when we call them good or 
bad, and he even agrees that moral judgments could be called "expressions of emotion, " 
since prescriptions are expressions, in a sense. But Hare cautions against taking 
Stevenson's idea too literally. In The Language of Morals, he writes: 

We speak of expressing statements, opinions, beliefs, mathematical relations, and so on; 
and if it is in one of these senses that the word is used, the theory, though it tells us little, 
is harmless enough. But unfortunately it is also used in ways which are unlike these; and 
Ayer's use (in speaking of moral judgements) of the word "evince" as its rough synonym 
was dangerous. Artists and composers and poets are said to express their own and our 
feelings; oaths are said to express anger; and dancing on the table may express joy. Thus 
to say that imperatives [or moral judgments] express wishes may lead the unwary to 
suppose that what happens when we use one, is this: we have welling up inside us a kind 
of longing, to which, when the pressure gets too great for us to bear, we give vent by saying 
an imperative [or moral] sentence. Such an interpretation, when applied to such sentences 
as "Supply and fit to door mortise dead latch and plastic knob furniture", is unplausible. 

In the 1980s and 1990s Simon Blackburn and Allan Gibbard developed versions of 
emotivism (or in Gibbard's more general terminology, "expressivism") grounded in the 
same root ideas as Stevenson's theory. These theories are more sophisticated in various 
ways (in particular they make good headway into the embedding problem mentioned 
above), and they have to some extent supplanted Stevenson's emotivism, though as 
inheritors, not as refuters. 


Works by Stevenson 

1944: Ethics and Language, New Haven, CT: Yale University Press. 

1963: Facts and Values: Studies in Ethical Analysis, New Haven, CT: Yale University Press. (A col- 
lection of papers. Essay Two is especially useful as an introduction. Essay Eleven contains the 
most mature version of the theory.) 

Works by other authors 

Ayer, A. J. (1936) Language, Truth and Logic, New York: Oxford University Press. (Chapter 6 con- 
tains the germ of emotivism, which Stevenson developed into a full and sophisticated theory.) 



Blackburn, S. (1984) Spreading the Word, Oxford: Oxford University Press. (Chapter 6 especially 
develops a variation on emotivism designed to address prominent objections.) 

Blanshard, B. (1949) "The New Subjectivism in Ethics," Philosophy and Phenomenological Research 
9, pp. 504-11. 

Geach, P. T. (1960) 'Ascriptivism," Philosophical Review 69, pp. 221-5. (An influential, but 
somewhat technical, objection to emotivism.) 

Gibbard, A. (1990) Wise Choices, Apt Feelings, Cambridge, MA: Harvard University Press. 

Goldman, A. and Kim, J. (eds.) (1978) Values and Morals: Essays in Honor of William Prankena, 
Charles Stevenson, and Richard Brandt, Dordrecht and Boston: Reidel. (A collection of critical 
essays; those on Stevenson's work are quite accessible. Contains a comprehensive bibliography 
of Stevenson's writing.) 

Hare, R. M. (1952) The Language of Morals, Oxford: Oxford University Press. (A work con- 
temporary with Stevenson's, with important similarities and contrasts.) 

Searle, J. (1969) Speech Acts, Cambridge: Cambridge University Press. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

W. V. Quine (1908-2000) 


Willard Van Orman Quine was born on June 25, 1908. He was graduated from Oberlin 
College with a degree in mathematics, summa cum laude, in 1930; his senior honors 
thesis was a proof within the system of Whitehead and Russell's Principia Mathematica 
(1910-13), which he studied largely without aid from his teachers. Whitehead was in 
the Philosophy Department at Harvard, so it was there that Quine went to do graduate 
work, although Whitehead was no longer teaching logic and Quine had done little 
undergraduate work in philosophy. Nonetheless, he completed a Ph.D. in two years, 
graduating in 1932. His dissertation generalized Principia 's treatment of classes so that 
it included dyadic relations, instead of treating the latter separately. Along the way, 
Quine clarified and reformulated the basis of the system - a point to which we shall 

Quine spent the academic year 1932-3 in Europe on a Sheldon Fellowship. He spent 
five months in Vienna, attending some meetings of the Vienna Circle. More important, 
perhaps, was a shorter stay in Prague, where he had extensive conversations with 
Rudolf Carnap, then completing The Logical Syntax of Language (Carnap 1934). While 
in Europe he was elected as one of the first group of Junior Fellows in Harvard's newly 
formed Society of Fellows: a three-year fellowship, without teaching obligations. He 
spent most of these years working on logic and set theory, though some of it on other 
aspects of philosophy. Under this latter head, he gave three lectures on Carnap, essen- 
tially expounding, in a strongly approving fashion, what he took to be the doctrines of 
The Logical Syntax of Language. (The text of these lectures is now published in Creath 
1990.) In 1936 Quine became Faculty Instructor at Harvard. Except for service in the 
US Navy during World War II, he held faculty positions there from that time until his 
retirement in 1978. He remained philosophically active and engaged for twenty years 
after retirement, continuing to write and publish into his nineties. 

It will be helpful to put Quine's work in the context of what I shall call twentieth- 
century scientific philosophy, a movement within the broader stream of twentieth- 
century analytic philosophy. Key figures in twentieth-century scientific philosophy 
(other than Quine) include Bertrand Russell and Rudolf Carnap, as well as others often 
identified as logical positivists or logical empiricists; Frege and Wittgenstein also made 
crucial contributions to the movement. 



Let us try briefly to characterize this movement by aims and doctrines rather than 
by its participants. Perhaps most notable is the emphasis on knowledge, and its objects, 
rather than on ethics or politics or aesthetics or history or the human condition, as the 
primary concern of philosophy; an emphasis, one might say, on the True rather than 
on the Good or the Beautiful. This emphasis is equally an emphasis on science, espe- 
cially on the natural sciences. It is characteristic of scientific philosophy to take the 
natural sciences as paradigmatic of all knowledge. Part of this view is the doctrine that 
the Vienna Circle called "the Unity of Science." The point here is the unity of all real 
knowledge, for the German word Wissenschaft is broader in its scope than most current 
uses of its invariable English translation, "science." (Ouine, however, makes it explicit 
that he uses the word "science" broadly; see Ouine 1995: 49.) According to this view, 
there are no fundamental divisions of aim or method among the various branches of 
knowledge. Along with this, there is a suspicion, or worse, of the claims of metaphysics, 
and of any claims neither answerable in straightforward fashion to the findings of 
empirical science nor provable by logic and mathematics. 

This brief sketch at once raises questions about there being any role for philosophy, 
even that of the scientific philosophers themselves. A central idea here was that phi- 
losophy is not to add to our knowledge but is, rather, to analyze the knowledge that the 
sciences give us, and thereby to give us greater clarity about that knowledge and its 
basis. The tool of this analysis was, above all, logic: the logic of Frege and Russell (see 
FREGE and RUSSELL). This logic held out an ideal of clarity; one aspect of the philoso- 
pher's task was to impose a similar clarity upon other subjects. This line of thought sug- 
gests an assimilation of philosophy to logic, but does not by itself account for the 
possibility of either of these subjects. Russell sought to do this by postulating an a priori 
insight, which might strike some as a large concession to metaphysics, in the pejora- 
tive sense. Carnap, drawing on the early work of Wittgenstein, held that logic is ana- 
lytic, empty of content, and hence not genuine knowledge at all. Philosophy too makes 
no claims about the world. Its analyses of language simply make explicit what is already 
there; it recommends a certain kind of language for this or that scientific purpose, but 
a recommendation is not a claim, and is presumably not in need of the same sort of 
justification. This emphasis on language is connected with Carnap's view that analytic 
truths are true in virtue of language, true by virtue of the meanings of the words 
making them up. Given the importance of analyticity as accounting for logic, for math- 
ematics, and for philosophy itself, this throws an enormous explanatory burden on the 
notion of language. 

We have just very briefly sketched the tradition of twentieth-century scientific philoso- 
phy. Quine's position relative to this tradition is ambivalent. On the one hand, he is 
its greatest exponent in the last forty years of the century. On the other hand he revo- 
lutionizes it, in such a way that one might say that he rejects the tradition rather than 
continuing it. Both Russell and Carnap attributed great importance to the natural 
sciences but nevertheless held that logic, mathematics, and philosophy itself, all have 
a status that is quite different from that of, say, physics or chemistry, or history or 
sociology. The former are independent of observation, and thus a priori - however 
exactly that idea is to be understood - while the latter are a posteriori, empirical 
and ultimately answerable to observation and sensory experience. Ouine rejects the 



idea that there is a fundamental epistemological distinction here. This rejection - which 
Quine himself sometimes speaks of as his naturalism - is fundamental for his philoso- 
phy. We need to see why he rejects the a priori, and how he can get by without it; 
we shall then to begin to show how his general approach to philosophy flows, in 
large part, from this step. 

Analyticity and the a priori 

Like Carnap, Quine rejects any idea of the a priori as based on pure intuition, or on pure 
reason; such an idea runs counter to his scientific and empiricist predilections. Carnap 
appealed to the idea of analyticity as an alternative (see CARNAP). Quine, famously, also 
rejects Carnap's use of this idea and with it any significant idea of the a priori or of 
necessity. We shall discuss Quine's arguments against Carnap's notion of analyticity, 
or against the idea that there is a serious and significant distinction to be made between 
the analytic and the synthetic. Quine's rejection of the distinction, however, is only half 
the story. The other half is to show how he can make sense of the apparently a priori 
status of logic and mathematics without it - or, better, perhaps, how he can account 
for those facts which have led philosophers to think that logic and mathematics must 
be a priori. 

Understanding Quine's attack on Carnap's notion of analyticity is complicated, 
partly because it was for a time, I think, not entirely clear even to Quine himself exactly 
what he is attacking and how. Quine thinks of Carnapian analyticity as truth in virtue 
of meaning, and so also thinks that if we had a clear understanding of the notion of 
meaning - more precisely, of synonymy, or sameness of meaning - then we would have 
gone a long way towards making clear sense of a notion of analyticity. So for a long 
time Quine's attack on analyticity seemed to be part and parcel of an attack on the 
notion of meaning, as unclear or undefined. And certainly Quine is skeptical as to how 
far we can make clear sense - which for him means sense in scientific, especially behav- 
ioral, terms - of the idea of synonymy. (He is not, however, skeptical of the notion of 
meaning/u/ness. See his essay "The Problem of Meaning in Linguistics," in Quine 1961, 
the burden of which is precisely that the notion of meaningfulness is not afflicted with 
the same sorts of problems as the notion of sameness of meaning, but lends itself to a 
ready, if somewhat rough, understanding in behavioral terms.) His view of synonymy 
is not that there is no sense at all to be made of it anywhere, even though he has reason 
to think that we may not be able to make complete sense of it everywhere. So we see 
Quine, as early as "Carnap and Logical Truth" (1963, written in 1954) accepting that 
there may be a limited notion of analyticity to be had. His willingness to accept some 
notion of analyticity becomes more marked as time goes by. In Roots of Reference ( 1 9 74) 
he proposes a tentative definition of synonymy, and with it an understanding of ana- 
lyticity; in "Two Dogmas in Retrospect" (1991) we find Quine arguing that (first-order) 
logic is analytic. How are we to understand this situation? 

Quine continues to reject the idea of a notion of analyticity that would play any- 
thing like the central philosophical role that Carnap allotted it. In order to play that 
role, a notion of analyticity would have to meet two requirements. First, it would have 
to have the right scope: the truths of logic and mathematics, at least, must come out 
as analytic. Second, it must also, at least in Quine's view, mark a significant epistemo- 



logical distinction: analytic truths need no justification (or else what counts as "justifi- 
cation" for them is wholly different in kind from the justification of synthetic truths). 
While Quine accepts a notion of analyticity, it is not one that satisfies either of these 
requirements. Let us begin with the question of scope and definition. 

Quine, as we saw, takes analyticity to be truth in virtue of meaning. But how are we 
to understand the idea of meaning, as it occurs here? For Quine, the only thing that 
could be relevant to the meaning of a word or a sentence in a given language is how it 
is used by speakers of that language. This is an important point. Quine has been accused 
of being unduly behavioristic, especially about language. Certainly he has a general 
bias in favour of a behaviorist approach to the mind. He claims, however, that his insis- 
tence on approaching language-use behaviorally is not merely the result of prejudice. 
Indeed he offers an argument for some form of behaviorism in this context. The passage 
is worth quoting at some length: 

In psychology one may or may not be a behaviorist but in linguistics one has no choice. 
Each of us learns his language by observing other people's verbal behavior and having his 
own faltering attempts observed and reinforced or corrected by others. We depend strictly 
on overt behavior in observable situations. As long as our command of our language fits 
all external checkpoints ... so long all is well. Our mental life between checkpoints is indif- 
ferent to our rating as a master of the language. There is nothing in linguistic meaning 
beyond what is to be gleaned from overt behavior in observable circumstances. (Quine 
1990: 37-8) 

For Quine there can be no more to meaning than is implicit in the actual use that is 
made of the language. 

Quine's interest is exclusively in knowledge, and the aspect of the use of language 
that primarily concerns him is our accepting or not accepting sentences. Thus he says, 
early on: "in point of meaning ... a word may be said to be determined to whatever 
extent the truth or falsehood of its contexts is determined" (1936: 89). But then the 
question is: which of the contexts of a word must be so determined in order to deter- 
mine its meaning? Without some reason to discriminate, we have no reason to treat 
one context as more definitive of a word's meaning than any other. But then no true 
sentence in which the word appears would have any better claim to be analytic than 
any other such sentence; clearly no useful analytic/synthetic distinction can be erected 
on that basis. 

What sort of thing might give us reason to discriminate among contexts? If mastery 
of some small subset of a word's uses gave one mastery of its use as a whole, then there 
would be reason to say that those uses, those contexts, constituted its meaning. And 
clearly this happens in some cases. A child who otherwise has a fair degree of linguis- 
tic sophistication but does not know the word "bachelor" can be given a mastery of that 
word all at once, at a single stroke, by being told that bachelors are unmarried men. 
This fact gives us every reason to say that "bachelor" means "unmarried man," and that 
the sentence 'All bachelors are unmarried" is analytic - which Quine, at least in his 
later work, certainly accepts (see Quine 1991: 270). Along these lines, he proposes a 
definition of analyticity: "a sentence is analytic if everybody learns that it is true by 
learning its words" (1974: 79). He argues that first-order logic is analytic by this sort 



of definition, but that other analytic truths will all be trivial. In particular, there is no 
prospect of arguing on this sort of basis that mathematics is analytic; apart from other 
considerations, Godel's incompleteness theorem would be an insurmountable barrier 
to such an argument (see TARSKI, CHURCH, GODEL). 

We have yet to discuss the question of the epistemological significance of the notion 
of analyticity This, Quine came to see, is the crucial question; in the 1980s he wrote: 
"I now perceive that the philosophically important question about analyticity and the 
linguistic doctrine of logical truth is not how to explicate them; it is the question of their 
relevance to epistemology" (Hahn and Schilpp 1998: 207). Why should anyone think 
that showing a sentence to be analytic for a given language - learned in the course of 
learning the language - shows anything about its epistemological status? Why might 
one think that it shows that for that sentence no justification is required, or that the 
question of justification is somehow misplaced? Well, clearly it might be thought to 
show that given that we are speaking that language the question of the justification of that 
particular sentence does not arise. But why does the question not simply become one 
of the justification for speaking that language? We are presumably operating here with 
very tight identity-criteria for languages, so that shifting the meaning of the one word 
"bachelor" would mean that we were speaking a different language (if this seems exces- 
sively odd, we might speak in terms of idiolects rather than languages; but the point is 
the same). And given that conception of a language, it is not obviously absurd to ask 
for the justification for speaking a given language (it is no longer enough to say, "it is 
the one f was brought up with and feel most at home in," for this quality would survive 
minor shifts). 

For Carnap, the choice of a language is in epistemologically important ways unlike 
the choice of a theory within a language. The former is not a matter of correctness, of 
right or wrong; it is a practical matter having to do with pragmatic factors such as the 
simplicity of a given language and its convenience for this or that goal. This idea issues 
in what he calls the "Principle of Tolerance": since choice of language, unlike the choice 
of a theory within a language, is not a matter of correctness or incorrectness, we should 
be tolerant, and allow people to work with whatever language they choose. Within a 
given language, justification is more or less rule-governed, governed by the rules of that 
particular language; justification, like other significant philosophical notions in 
Carnap's view, is thus language-relative. But the choice of a language itself is not some- 
thing that can be justified in the same sort of way, since without a language we have 
no rules of justification to which to appeal. 

Carnap emphasizes the distinction between the justification of choice of a theory 
within a language and the justification (or the lack of need for justification) of choice 
of language. The idea that attributing analyticity to a sentence has epistemological sig- 
nificance depends upon this distinction. Saying of a sentence that it is analytic would 
mean that it is in some sense integral to the language that we currently speak. So if we 
ceased to accept that sentence we would have modified the language. But that would 
leave open the possibility that we might have evidence which would justify that modi- 
fication of the language. Thus it would seem that evidence might bear on an analytic 
sentence in the same sort of way in which it bears on a synthetic sentence, unless the 
notions of evidence and justification in play are of different kinds in the two cases. 
Carnap, of course, holds there there is just such a difference in kind. He claims that 



there are quite different conceptions of evidence and justification at work. Within the 
language, justification is a rule-governed procedure, and a matter of right or wrong; 
when the language itself is being chosen, however, there are no rules to which to 
appeal, and the choice is purpose-relative and to some extent arbitrary. 

Quine attacks this distinction from both sides. He denies that (internal) justification 
is to any significant extent a rule-governed procedure. He says, for example: 

I am impressed . . . apart from prefabricated examples of black and white balls in an urn, 
with how baffling the problem has always been of arriving at any explicit theory of the 
empirical confirmation of a synthetic statement. (1961: 41-2) 

He also insists that all our cognitive choices, including the choice of a language for 
knowledge, are directed towards the same end: achieving the most successful theory, 
where a crucial test of success is the generation of true predictions. Vaguer virtues, 
such as simplicity and fruitfulness are also relevant. These are the sorts of things that 
Carnap counted as "pragmatic factors," applicable to questions of language-choice. 
Quine claims that they are applicable also to what Carnap would count as empirical 
beliefs. They may not in any very obvious way be applicable to the question whether 
there is now a desk in front of me, but certainly they are to more or less abstract claims 
of theoretical physics. For Quine there is a continuum here, with no sharp breaks 
to be had. 

Quine thus holds that even where we have a significant truth which is analytic, this 
status simply does not matter epistemologically: 

"Momentum is proportional to velocity" counts as analytic. But do we care? Einstein's 
relativity theory denies the proportionality law, complicating it with a formula involving 
the speed of light. But instead of accusing Einstein of a contradiction in terms, we simply 
stand corrected. (Quine 2000) 

Now Carnap might agree that we "stand corrected" because we accept that Einstein 
has shown us that a non-Newtonian language works better for making some predic- 
tions, but he would insist that this is a different sense of correction from that in which 
we are corrected when we change our mind about a belief that does not involve a 
change of language. But this is precisely what Quine denies, as we have seen. 

One issue which arose above was Quine's view of the nature of justification, the rela- 
tion between the evidence we have and the beliefs that we hold on the basis of it. The 
point there was that there is not, in general, a simple relationship between a sentence, 
on the one hand, and an observation or group of observations that justify it, on the 
other hand. Justification is not, in general, a simple and rule-governed matter. Of course 
there are sentences, such as "there is a desk in front of me now," which do seem to have 
a very straightforward relation to observations. What makes that case straightforward 
is that it hardly matters what else a person believes: given the right observations, almost 
anyone will accept that there is currently a desk in front of them. The justification rela- 
tion here holds between observations and the individual sentence believed, whatever 
one's other beliefs may be. fn Quine's view, however, this is a poor paradigm to use for 
knowledge as a whole. In general, justification is holistic, meaning that it does not apply 



to sentences taken individually, in isolation from others, but rather to larger or smaller 
chunks of theory, made up, in some cases, of a large number of sentences. Many of the 
sentences we accept - most obviously the more abstract and theoretical ones - have 
relations to observations only if we tacitly assume many other sentences. These other 
sentences, background assumptions, are required if the sentence in which we are inter- 
ested is to have any observational consequences at all. From the point of view of the 
working scientist, the background assumptions may be confidently accepted, and only 
the individual sentence up for testing. From a more abstract point of view, such as 
Quine's, however, what is tested by observation is not the individual sentence alone, but 
rather the whole set of sentences that implies the observational consequences. From a 
sufficiently abstract point of view, indeed, it is always the whole of our knowledge that 
is tested. Any test of a sentence presupposes truths of logic among its background 
assumptions. Logic, however, is used everywhere in our system of beliefs, so in a rather 
Pickwickian sense it is that system as a whole that is at stake. (Quine calls this extreme 
holism "legalistic" (1991: 268). I take this to mean that it holds from a very abstract 
point of view, but not that it is unimportant.) 

Holism is not new with Quine. When "Two Dogmas" was reprinted, Quine added a 
footnote to Duhem, and there is a clear statement of the view in Carnap's Logical Syntax 
of Language, with references to Poincare as well as to Duhem (Carnap 193 7: 318). 
Quine's uses of the doctrine, however, are novel. One use we have seen: it is the basis 
of the claim that justification (within a language) is not the sort of rule-governed pro- 
cedure that Carnap sometimes suggests, and so is not different in kind from the sort of 
justification that applies to the choice of one language rather than another. This claim, 
in turn, is crucial for Quine's view of analyticity and the a priori, discussed in the pre- 
vious few pages. 

A second use that Quine makes of holism also relates to the question of the a priori, 
but in a quite different way. He attempts to undercut the idea that there must be a priori 
knowledge by invoking holism to explain the phenomena which led some philosophers 
to invoke the idea of the a priori, but to do so without invoking that idea. (If those phe- 
nomena can indeed be explained without the a priori, then their existence no longer 
constitutes a reason to accept the a priori.) There is no doubt that the theorems of math- 
ematics and of logic are not discovered by experiment or, at least in any ordinary sense, 
justified by observation. And their falsehood seems completely inconceivable. How can 
these facts be explained in accordance with Quine's views? Holism provides the answer. 
Logic figures everywhere in our system of beliefs; mathematics is used in many 
branches of knowledge. No one observation or experiment bears on them, but the 
success of our system of beliefs as a whole in predicting experience provides justifica- 
tion; this justification is exceedingly indirect. For Quine that only puts it at one end of 
a continuum which we already have reason to accept, for "e = mc 2 " is justified very 
much less directly than is "there is a desk in front of me now." Equally the unimagin- 
ability of the falsity of logic becomes intelligible. Given the ubiquity of logic, changing 
it means making changes everywhere in our system of knowledge. It is not to be 
wondered at if this is hard to conceive. 

Our discussion so far has been focused on Quine's rejection of the a priori. He is thus 
left with no kind of knowledge other than the ordinary knowledge of common sense 



and (better) science. Philosophy too, since it claims to yield knowledge, must be of this 
same general sort. This is the doctrine that Quine calls naturalism; it is absolutely fun- 
damental to his thought. The rest of this essay will take naturalism as its starting point, 
and investigate Quine's philosophy as an unfolding of that doctrine. I shall begin with 
topics having to do with epistemology and then move to topics whose focus is ontology 
and metaphysics (or its Quinean analogue). There is, however, one further issue which 
we should briefly mention here. 

Quine has argued for the possibility that two translators might come up with differ- 
ent translations of some sentences - not merely stylistically different, but "not equiva- 
lent in any plausible sense of equivalence, however loose" (Quine 1960: 27). This is the 
controversial doctrine known as the indeterminacy of translation. Where sentences are 
not asserted or denied on any very direct observational basis, the evidence (behavioral 
evidence, of course) for one translation over another is mediated by other sentences; 
alternative translations of all of them might cancel out, leaving each of two overall 
schemes of translation as equally justified. In Quine's view there would, in such a case, 
be no right and wrong, no fact of the matter: it would not be our knowledge which was 
lacking, but rather that there was no fact to be known. This idea provoked an enormous 
amount of discussion in the 1960s and early 1970s, and some commentators have 
even thought that it is what really underlies Quine's objections to analyticity In my 
view, however, it is not of great importance to his thought, except as dramatizing the 
idea that meaning must ultimately be answerable to behavior (see Hylton 1990). Note 
in this connection that Quine now speaks of indeterminacy as "a conjecture" (see Hahn 
andSchilpp 1986: 728). 

Knowledge and the realm of the cognitive 

How should we conceive of knowledge, if we are to take a scientific approach to it? 
Fundamental to Quine's thought is the idea that knowledge is to be understood as a 
biological phenomenon. Human knowledge is thought of as a condition of the human 
animal. It originates in the struggle of one species of primate to survive. The opening 
sentences of From Stimulus to Science read like this: "We and other animals notice what 
goes on around us. This helps us by suggesting what we might expect and even prevent, 
and thus fosters survival" (Quine 1995: 1). This is what one might call a Darwinian 
conception of knowledge: knowledge as an adaptive mechanism, fostering the survival 
of the species. 

In this view we see a decisive rejection of the long-standing philosophical tradition 
that sharply distinguishes between real knowledge and mere belief or opinion, between 
scientia and doxa. That tradition tends to assimilate real knowledge - scientia - to knowl- 
edge of a proposition of mathematics, when it is known on the basis of a thoroughly 
understood proof. Real knowledge is accordingly thought of as infallible and known 
with certainty. None of these ideas fits with the idea of knowledge as a biological phe- 
nomenon, as what helps the human animal to get by in its dealings with the world. 
Some philosophers have held that those ideas are implicit, more or less, in the word 
"knowledge." Quine's response is simply to abandon that word for "scientific and philo- 
sophical purposes" (1987: 109). He continues to use it informally (as shall I in 
expounding his views), but without the weight that it may be thought to carry in 



general use. He makes no sharp distinction between our knowledge, and that which 
we, or experts among us, accept upon reflection. Quine's conception of cognition, then, 
is fallibilist through and through: no part of our system of beliefs can be counted as 
wholly immune from revision, though some parts are no doubt far more secure than 

A second very general point about Quine's conception of knowledge is that he takes 
it to be linguistic or verbal; at least for "scientific and philosophical purposes," he thinks 
of our system of beliefs as being embodied in sentences: 

What sort of thing is a scientific theory? It is an idea, one might naturally say, or a complex 
of ideas. But the most practical way of coming to grips with ideas, and usually the only 
way, is by way of the words that express them. What to look for in the way of theories, 
then, are the sentences that express them. (Ouine 1981: 24) 

Much of Quine's interest in language and in its analysis arises from the fact that our 
knowledge is embodied in language. 

In one way, as we saw, the idea of knowledge, as Ouine employs it, marks no very 
sharp distinction. Unlike some philosophers, he does not use the term as an honorific, 
connoting some particular high degree of justification or of certainty. He does, however, 
make a sharp distinction between the realm of the cognitive and the rest of human 
activity. To call a human activity cognitive is, roughly, to say that it is answerable to, if 
not exclusively aiming at, predictions of sensory experience. It is perhaps a presuppo- 
sition here that cognitive activity and cognitive language can be peeled off from the 
chaotic mass of human activity and language generally - or at least that we can 
abstract without distortion, and talk of the cognitive while ignoring the rest. 

The prediction of experience is a practical matter; at the limit, as we indicated, 
survival is at stake. This is the sense in which Quine's conception of knowledge is 
Darwinian. Thus far Ouine is with the pragmatists. On the other hand, tying the 
concept of knowledge to the prediction of sense-experience enables Ouine to make clear 
distinctions, and to erect barriers, in places where the pragmatists would not. Activities 
which predict experience are cognitive; others, though they may contribute to human 
flourishing in other ways, are not. The justification for this distinction is presumably 
(for Ouine is not explicit here) that sense-experience is our only way of finding out about 
the world (we shall discuss this idea shortly). In spite of Quine's practical, Darwinian 
view of knowledge, there is thus a sense in which his view makes a clear distinction 
between the theoretical and the practical. Theoretical success is success in prediction of 
sensory experience. 


As traditionally conceived by philosophers, this notion includes two strands. On the one 
hand, evidence is thought of as consisting of the epistemologically most fundamental 
items of our knowledge: evidence is that which is, so to speak, first in the order of 
knowledge (that which is evident), and so also that from which other items of knowl- 
edge must be inferred. On the other hand, evidence is also to consist of immediately 
given data, devoid of any conceptual impositions of our own; since no interpretation is 



involved, there is no room for doubt. These two strands are in tension. What is literally 
first in the order of knowledge seem to be facts about other people and ordinary physi- 
cal objects. Yet sentences recording such facts do not seem simply to record raw data. 
They involve conceptualization, and are (notoriously) open to doubt. 

Quine's response to these difficulties is to abandon the traditional conception of evi- 
dence completely, in favour of a physicalistic alternative. He speaks not of "the given" 
but of the stimulation of our sensory receptors, and of observation sentences: roughly, 
sentences that any speaker of the language is disposed to accept or reject simply on the 
basis of current stimulation. Thus he says: 

our immediate input from the external world [is] the triggering of our sensory receptors. 
I have cut through all this [i.e. the difficulties of analyzing the notions of observation and 
experience] by settling for the triggering or stimulation itself and hence speaking, oddly 
perhaps, of the prediction of stimulation . . . 

Observation drops out as a technical notion. So does evidence, if that was observation. 
We can deal with the question of evidence for science without the help of "evidence" 
as a technical term. We can make do instead with the notion of observation sentence. 

What is at stake in philosophical talk of evidence is, from Quine's point of view, the 
issue of how we find out about the world. Here we have a crucial example of his method, 
of the idea that the study of knowledge is to be naturalized, and as far as possible put 
on a physicalistic basis. For him, the issue is to be taken as a scientific question: how do 
we come by information about the world? And the answer is that we do so by the impact 
of various forms of energy on our sensory surfaces. Physics will tell us what forms of 
energy there are; physiology and psychology will say which forms of energy human 
beings can detect, i.e. to which forms human beings respond. The central fact here is 
that it is only through stimulation of our nerve-endings by energy impinging on our 
sensory surfaces that we human beings know anything at all about the world. This is 
fundamental to Quine's epistemology and it is, he emphasizes, "a finding of natural 
science itself " (1990: 19). 

Quine's use of the notion of stimulations is thus symptomatic of his general shift in 
perspective. The question is not: what is given to me, at the outset of my cognitive 
endeavors? But rather: how do we humans gain knowledge of the world? This is on his 
view a straightforward scientific and causal question. The answer refers us to stimula- 
tions of our sensory surfaces. Quine thus abandons, or greatly modifies, the traditional 
concept of sensory evidence, fn so doing he shifts the question to which the original 
conception was an answer. The traditional philosopher's demands are not met by 
Quine's view. In the most straightforward sense, the occurrence of stimulations is inde- 
pendent of and prior to theory. Our knowledge of such matters, however, is clearly not 
independent of theory. So there is here no prospect of what the traditional philosopher 
sought: support of the theory which is wholly independent of the theory. To the extent 
that this strikes us a problem or a paradox, to that extent we have not accepted Quinean 

The point we have just noted links Quine's rejection of the sensory given with his 
rejection of the traditional conception of the a priori, or of any analogue. The given 



and the a priori were each conceived, by some philosophers, as a kind of extra- 
theoretical knowledge, knowledge somehow free of the vicissitudes affecting the 
ordinary knowledge of common sense and science. Quine denies that there is any such 
kind of knowledge. Within what we take ourselves to know there is, no doubt, better 
and worse; there is, however, no knowledge of a wholly different and superior kind. 
Quine is consistent here. He does not take even the most fundamental points of his own 
philosophy to be a priori - and this includes the doctrine that we know about the exter- 
nal world through impacts on our sensory surfaces. While it is extremely unlikely, there 
are imaginable circumstances under which we might drop our present idea of sensory 
evidence entirely. If sufficient confusion resulted from our following the evidence of our 
senses, as we now understand them, and sufficient success on the part of those who 
claim to hear voices in their heads, say, our estimate of the role of stimulation of our 
sensory surfaces might change. (This point is explicit; see Quine 1990: 20f.) 

For Quine, as we have said, the central fact about knowledge is that it is only through 
stimulation of our nerve-endings that we know anything at all about the world. Such 
stimulations provide the only empirical constraint on our system of beliefs, the only 
external criterion of success. (By speaking of an external criterion I mean to leave room 
for what one might think of as internal factors: such as the overall simplicity of the 
system.) This fact suggests that there may be empirical slack between evidence, even 
the totality of all possible evidence, and theory. We can focus this idea by asking 
whether there might be two systems of belief, different from one another but each fully 
successful at "predicting stimulations." Quine's answer, though somewhat qualified 
and complicated, is that nothing rules this out. This is the doctrine known as the under- 
determination of theory by evidence. Even if, per impossibile, all the evidence were in, still 
any given theory based on that evidence might not be uniquely justified. If we had a 
theory that explained and predicted the evidence satisfactorily, that would of course 
justify it; since it would at least in principle be possible for another theory to do as well, 
however, the justification would not be unique. 

The relation of evidence to knowledge: 
observation sentences 

So far we have said something about evidence and something about knowledge, as 
Quine conceives them, but nothing explicit about the relation between the two. Clearly 
there is a gulf between the sensory stimulations that are our only source of informa- 
tion about the world, and the mass of sentences in which our beliefs about the world 
are embodied. How is this gulf bridged? For the sentences most directly tied to sensory 
evidence, observation sentences, there is a fairly clear account. The other sentences of 
our theory of the world get their relation to evidence via their relation to observation 
sentences, and here the account is much sketchier. (A full account is, Quine thinks, not 
yet available, and may never be.) Discussing observation sentences will at least give us 
an idea of how the gap between evidence and theory is bridged. 

The first point to make about what Quine calls observation sentences is that they 
are - by his lights - sentences. This does not mean that they are all sentences in the 
grammatical sense: "Rabbit," taken as a complete utterance, is Quine's own example. 
The idea of a sentence here is that of a piece of language which may be used to 



say something, and thus may be true or false. For Quine, the fundamental evidential 
relation holds between sensory stimulations and units of language of that sort. 

We can bring out the significance of this idea by contrasting sentences (as we are 
using that word) with referring expressions, or "terms," as Quine says. In most uses, 
the word "rabbit" is a term, not a sentence: if I say "I see a rabbit" then in that sentence 
it is a term which functions as part, but only part, of a sentence. Mastery of the word 
in that sort of use, Quine claims, requires more than a mere ability to respond to the 
presence of rabbits, ft requires also that one can distinguish the circumstances that 
license the claim "There's one rabbit" from those licensing the claim "There are many 
rabbits"; that one can similarly distinguish "There's the same rabbit again" from 
"There's another rabbit"; and so on. The ability to make these and related distinctions 
requires some knowledge of the ways of physical objects in general, and of rabbits in 
particular. Sentences containing terms are thus not epistemologically basic. Mastery of 
the use of a term already requires that one possess some knowledge, and then the ques- 
tion of the evidence for that knowledge must arise. Mastery of the word "rabbit" as an 
observation sentence, by contrast, requires only the capacity to respond to environ- 
ments containing rabbits in ways in which one does not respond to environments which 
do not contain rabbits; this capacity requires no auxiliary knowledge. This kind of 
ability is all that is presupposed by the mastery of an observation sentence. It is primi- 
tive and fundamental. It is what underlies all cognitive language-use, including the 
most sophisticated, but other forms of language go beyond it in principle. 

To this point we have said little more about observation sentences than that they are 
sentences, though that idea has proved to be far from trivial. Beyond that, an observa- 
tion sentence is what Quine calls an "occasion sentence," i.e. it is one that is true when 
uttered under some circumstances and false under others (so "It's raining in London" 
is an occasion sentence; "Gold is a metal" is not). The rough idea that Quine intends to 
capture is that an observation sentence is an occasion sentence about which there is 
community-wide agreement under any given circumstances. A little more precisely, we 
can distinguish two conditions which an occasion sentence must satisfy to count as an 
observation sentence. First, whether a given speaker of the language is disposed to 
assent to, or dissent from, an observation sentence at a given time is simply a matter of 
the stimulations that that individual is undergoing at that time. For each individual, 
the same stimulation-pattern will typically lead to the same verdict each time. Second, 
any fully competent speakers of the same language, in the same circumstances, will 
agree on an observation sentence; Quine speaks of "unhesitating concurrence by all 
qualified witnesses" (1995: 44). (There is some vagueness here, arising from the vague- 
ness of "in the same circumstances," and of "qualified witness.") 

Observation sentences thus assert the presence of something readily detectable by 
the senses. Whether such a sentence is correctly assertable in a given situation does not 
depend upon ancillary information, unless it is shared by all speakers of the language. 
Hence "It's cold here!" might qualify, but "That's Quine!" would not, since not all 
English-speakers would recognize that philosopher on sight. The range of observation 
sentences will vary with our decision as to exactly who should be included among the 
fully-functioning speakers of the language. "That's red" will presumably count if we 
exclude the blind and the color-blind, but otherwise not. Observation sentences are the 
epistemologically most basic parts of our theory of the world. They can be known before 



anything else. (They thus play one of the roles traditionally accorded to the notion of 
evidence: they are first in the order of knowledge.) The knowledge that they embody is 
so rudimentary that in almost all cases it goes unspoken, but can be elicited by raising 
the sentence as a question and noting the subject's reaction. 

Considered holophrastically, i.e. as unanalyzed wholes, observation sentences are 
simply responses to stimulation, and are in only the most minimal sense conceptual or 
theoretical. Hence such sentences can be mastered by a child otherwise quite innocent 
of language. As Quine says, "Their direct association with current stimulation is essen- 
tial if the child is to acquire them without prior language" (1990: 5). Considered as 
made up of parts, however, they connect with sophisticated theory, for the words which 
make them up recur in more theoretical contexts. This dual aspect is essential to the 
function of observation sentences as the starting point of language and conceptual- 
ization. Because they presuppose so little, observation sentences will be the first sen- 
tences learned by a child (or, indeed, by an adult trying to find his or her unaided way 
in a wholly strange linguistic community); their learning presupposes no prior con- 
ceptual or theoretical resources. Because their terms recur in higher theory, learning 
such a sentence is a start on learning the language as a whole; it is only this sharing 
of vocabulary which unites the observation sentences with the rest of the language 
(1990: 8). 

Observation sentences, we saw, are occasion sentences: the truth-value of such a 
sentence will vary from one occasion of utterance to another. Our scientific theories, 
however, and most other serious knowledge, consists of standing sentences: sentences 
true or false once-for-all. How is this gap in turn to be bridged? Quine's answer appeals 
to the notion of an observation categorical. This is a sentence compounded of two obser- 
vation sentences, saying that whenever one of them holds the other will also hold: 
"Whenever there is smoke there will be fire," for example. This is a standing sentence, 
and so might be implied by a serious branch of organized knowledge. On the other 
hand, both of its component parts are observational (or so we are supposing). So we 
can tell right off whether a given situation is one in which there is smoke, and whether 
it is one in which there is fire. Hence we can tell right off whether a given situation is 
one in which the observation categorical is falsified. Because the sentence is in effect a 
generalization over all situations, it cannot, of course, be verified by a single situation, 
but it - and hence the theory which implies it - can be falsified. Quine readily accepts 
this asymmetry between verification and falsification, which fits with his general 
fallibilism. Since our theories have infinitely many observational consequences, they 
cannot be conclusively verified; in principle, however, a single observation may falsify 
a theory. 

Naturalized epistemology and normativity 

We have been articulating Quine's general conception of knowledge, evidence, and the 
relation between them. This conception is thoroughly naturalistic: Quine treats knowl- 
edge as a natural phenomenon, to be studied by the procedures of science. Most of 
Quine's own work in epistemology is an articulation and defense of this very general 
conception. He suggests, however, that there is also room in epistemology for detailed 
piecemeal work of a more recognizably scientific kind. This would consist in tracing the 



connections between theory and evidence in a psychologically realistic fashion, to see 
how our knowledge is in fact related to the evidence that we have. Epistemology of this 
sort is thus a branch of psychology. (Of course psychology is itself among the items of 
knowledge whose relation to evidence is to be investigated in this way; Quine speaks 
here of "reciprocal containment" (1969: 83).) 

Epistemology, as traditionally thought of, is a normative subject: it aims to tell us not 
merely about what is but also about what ought to be; it aims to tell us not only what 
we do in fact believe, and on what evidence, but also which beliefs are justified on the 
basis of the evidence that we have. Quinean epistemology, at least according to his 
account of the matter, is descriptive. To what extent, if any, can this descriptive subject 
take on the burden of traditional normative epistemology? This is a large and complex 
question (see Gregory 1999, to which I am indebted here). Roughly we may say that 
Quine has no room for the very large-scale questions and doubts which are one kind of 
starting point for traditional epistemology. He has no sympathy at all, for example, with 
global skepticism. The aim of our knowledge is to predict sense-experience. A theory 
that does that satisfactorily does all that we can ask. There is no further question as to 
whether it tells us about the nature of reality: 

what if ... we have achieved a theory that is conformable to every possible observation, 
past and future? In what sense could the world then be said to deviate from what the theory 
claims? Clearly in none. (Quine 1981: 22) 

We cannot divorce the idea of reality from that sense-experience. (This point will 
emerge further in the next section, below.) 

There is, nevertheless, a sense in which Quinean epistemology is normative. The 
criterion of success for all putative knowledge, for science in Quine's broad sense, is 
the prediction of sense-experience. Quine sees this as defining the notion of science 
(1990: 20). This definition is not arbitrary: our primary aim, in science, is to find 
out about the world, and one thing we know - a well-established piece of scientific 
knowledge - is that it is only through sense-experience that we come to know 
about the world. Given that our goal is fixed, there are questions of a normative 
sort about the best ways in which to achieve that aim, and in this instrumental 
sense epistemology is normative: 

Naturalization of epistemology does not jettison the normative and settle for the indis- 
criminate description of ongoing procedures. For me normative epistemology is a branch 
of engineering. . . . There is no question here of ultimate value, as in morals; it is a ques- 
tion of efficacy for an ulterior end, truth or prediction. The normative here, as elsewhere 
in engineering, becomes descriptive when the terminal parameter is expressed. (Hahn and 
Schilpp 1998: 664-5) 


As we have seen, Quine's naturalism can be identified with the view that there is 
essentially only one kind of knowledge. In particular, there is no special philosophical 
perspective from which we can attain knowledge that is independent of our 



ordinary scientific or commonsensical theory of the world. Thus on his account we 
are always inside that theory, modifying it, perhaps, but not wholly transcending 
it. There is no transcendental standpoint that is independent of our ordinary 
knowledge, and from which we can evaluate that knowledge without presupposing 
it. Quine's realism is an important application, and illustration, of this view. 

Quine's work is full of remarks which might suggest that he does not take our 
theories - including the "theory" that is common sense knowledge - to be (really) 
true, and does not take the objects that those theories presuppose to be (really) 
real. He says that our theories far outrun the evidence that we have for them, and 
that more than one theory is compatible with that evidence. The disparity between 
our evidence and our knowledge is a recurrent theme in his work. His insistence on 
this disparity, together with his view that our knowledge is justified by its efficacy 
in predicting and understanding the course of experience, might lead one to suppose 
that his is an instrumentalist or pragmatist view: that our theories, even at their 
best, are not really true, that they do not aim to correspond to an extra-theoretical 
world, but are simply useful instruments for predicting, understanding, and controlling 
future experience. Some critics have taken this to be Quine's view, and have seen 
his insistence on realism as a contradiction, or as a mis-statement of his actual 
position (see, for example, Lee 1986, and Smart 1969). Quine, however, insists that 
his view is "robust realism" (1990: 21), and that appearances to the contrary can be 
dispelled by taking naturalism seriously enough. 

On Quine's view, the objects that our theories presuppose do exist in extra-theoretic 
reality. It is part of our theory - that is, part of the best understanding of the world we 
have - that those objects (with a few exceptions, most of them straightforward) are not 
dependent on us or our theorizing. Now the critic may protest: it may be part of our 
theory that our theoretical objects really exist, independent of our theory - but that's 
just part of the theory. Do the objects really exist? But this is an attempt to ask a ques- 
tion from a stance independent of our theorizing about the world: the point of Quine's 
naturalism is that there is no such stance. The objects that we believe in exist, and are 
real, in the only sense of those ideas that we actually have. 

More generally, the apparently skeptical remarks that Quine makes when discussing 
our acquisition of knowledge do not affect his belief in the truth of the knowledge that 
is thus acquired: his ontology is in this respect insulated from his epistemology. This is 
a consequence of Quine's version of naturalism, and in particular of the reciprocal con- 
tainment of science, with its ontological claims, within epistemology, and vice versa. 
Let us see how this goes. We accept, let us suppose, the best overall theory of the world 
that is available. Freely drawing on this theory we do epistemology, i.e. we investigate 
the way in which human beings - including ourselves - come to formulate theories and 
to posit the existence of objects; among the theories thus investigated are those that we 
are drawing on in the course of our investigation. Now the crucial point is that to call 
a body of knowledge a "theory," or to call an object a "posit," does not in the least 
impugn its truth or its reality. "Theory" here is not mere theory, contrasted with real 
knowledge, for any body of knowledge will count as a theory from the point of view of 
epistemology; nor is "posited object" contrasted with real object. Thus, as Quine 
famously says: 



To call a posit a posit is not to patronize it . . . Everything to which we concede existence 
is a posit from the standpoint of a description of the theory-building process, and simul- 
taneously real from the standpoint of the theory that is being built. Nor let us look down on 
the standpoint of the theory as make-believe; for we can never do better than occupy the stand- 
point of some theory or other, the best we can muster at the time. (1960: 22; my emphasis). 

The crucial point is the one emphasized: there is no alternative to occupying some sub- 
stantive theory of the world, and to do this means accepting that theory, at least for the 
moment, as true, and accepting its objects as real. Of course we may develop our theory 
into a different one, but we cannot occupy some neutral philosophical vantage point; 
nor can we accept a theory while still pretending that we are not accepting it as true. 
In Quine's use, then, theoretical knowledge is not contrasted with ordinary know- 
ledge. Similarly theory for Quine is not contrasted with fact. All knowledge is theoreti- 
cal, in Quine's sense. Just as there is here a stretching (or a distortion) of the word 
"theory" so also there is a stretching of the word "posit." When we say of, neutrinos, 
for example, that they are posits, we would generally be taken to mean that some person 
or group of people consciously posited them. Quine, however, speaks of physical objects 
in general as posits, and here there is no such implication. No conscious decision was 
ever taken to posit such things. As in the case of "theory," Quine's use thus assimilates 
ideas which one might suppose to be importantly different. 

Metaphysics and regimentation: logic and extensionality 

To this point we have seen Quine reflecting on the nature of our knowledge, its sources 
and bases, and on its status. The philosopher's task, as he conceives it, also includes 
clarifying our knowledge and helping us to attain a clear view of just what it comes to 
and what it really commits us to. This latter kind of task may be thought of as the 
Quinean version or analogue of metaphysics. Like the others, it aims, in Quine's view, 
to contribute to the overall scientific enterprise, ft should, for example, enable us to 
avoid useless or misleading questions, help to suggest fruitful lines of further inquiry, 
and expose potential problems that may lurk in scientific theories. 

How is this task to be approached? The method here is to show how various parts 
of our knowledge could be reformulated in the clearest possible terms. We have 
reason to take the objects and categories revealed by this reformulation as real and 
fundamental. An extended passage from Word and Object is worth quoting at length on 
this topic: 

The same motives that impel scientists to seek ever simpler and clearer theories adequate 
to the subject matter of their special sciences are motives for simplification and clarifica- 
tion of the broader framework shared by all the sciences. Here the objective is called philo- 
sophical, because of the breadth of the framework concerned; but the motivation is the 
same. The quest of a simplest, clearest overall pattern of canonical notation is not to be 
distinguished from a quest of ultimate categories, a limning of the most general traits of 
reality. Nor let it be retorted that such constructions are conventional affairs not dictated 
by reality; for may not the same be said of a physical theory. True, such is the nature of 
reality that one physical theory will get us around better than another; but similarly for 
canonical notation. (1960: 161) 



Themes that we examined in the first part of this essay, having to do especially with the 
rejection of the Principle of Tolerance and with the remoteness of some parts of our 
theory from experience, re-emerge here. Simplicity, clarity, and convenience are not 
merely "pragmatic" virtues which are to be distinguished from the cognitive or theo- 
retical virtue of truth. If we have a language which displays those virtues to the highest 
extent, then we have reason to take the structure of that language as telling us some- 
thing about the real world, just as we take theories which display those virtues to tell 
us something about the world. 

One aspect of the philosopher's task, then, is to find the clearest and simplest frame- 
work in which to formulate our knowledge. In Ouine's view, this clarity and simplicity 
is to be found in first-order logic. (By "first-order" logic is meant logic of truth-functions 
and quantifiers which bind variables in positions occupied by singular terms, but not 
positions occupied by predicates.) He takes it that our knowledge is at its clearest when 
it is formulated in the syntax of first-order logic: a syntax which uses only truth- 
functions, predicates, variables, and quantifiers. This is not to say that he advocates lan- 
guage reform, or that he thinks that we should in fact reformulate all of our knowledge 
in those terms. But where our concern is with getting clear about what some part of 
our knowledge really commits us to, we would do well to consider how it might be 
phrased in logical syntax. This syntax is extraordinarily transparent and economical, 
which makes theorizing about it simple, and yet has surprising expressive power. It con- 
tains neither proper names nor function-symbols among its primitive expressions, for 
example, yet the effect of each can be easily achieved. (The basic techniques of doing 
so derive from Russell's Theory of Descriptions (see RUSSELL).) One moral to be drawn 
here is that we may be able to achieve the effect of a particular construction without 
in fact having to augment our stock of primitives. Quine is much concerned to take 
advantage of this kind of economy wherever it is possible, for the sake of the clarity and 
simplicity of the overall theory. (In ontology, as we shall see, Quine has a similar 
concern with economy: to show how particular kinds of object which appear to be 
assumed in what we take ourselves to know need not in fact be taken for granted as 
primitive, because their effects can be duplicated by other means.) 

One feature of this framework which has proved extremely controversial is its exten- 
sionality. A language is extensional when any sentence of it retains its truth-value 
under any one (or more) of three kinds of changes: (1) any name in it may be replaced 
by any other name of the same object; (2) any predicate in it may be replaced by a co- 
extensive predicate (i.e. one true and false of exactly the same objects); (3) any sentence 
embedded in another sentence may be replaced by a sentence of the same truth-value. 
It is important to note that these three requirements which a language must satisfy to 
count as extensional interlock in ways that make it very hard for a reasonably com- 
prehensive language to satisfy any one of them without satisfying all three. This point 
is important because the first requirement is, on the face of it, far more plausible than 
the third. It is extremely plausible that if we really understand what a given sentence 
is about, then we could replace the name of that object in the sentence with another 
name of the same object without altering the truth-value of the original. If a sentence 
is genuinely about an object, how we name that object should be a matter of indiffer- 
ence to the truth or falsehood of the sentence, though it may affect, for example, its 
poetic quality. (One might, indeed, take this idea as a partial definition of the somewhat 



vague idea of aboutness.) No such superficial plausibility attaches to the third require- 
ment, yet it turns out to be very hard to see how one can accept the first without also 
accepting the third. (See, for example, Quine 1995: 91-2.) 

Any language which uses the syntax of first-order logic (along with any standard 
semantics) is, in virtue of that fact, an extensional language. Yet it would be a mistake 
to think Quine accepts extensionality simply in order to be able to use the syntax of 
logic. On the contrary: he takes one of the advantages of that syntax to be that it 
enforces extensionality, which he holds to be desirable for its own sake: 

I find extensionality necessary . . . though not sufficient, for my full understanding of a 
theory. In particular, it is an affront to common sense to see a true sentence go false when 
a singular term in it is supplanted by another that names the same thing. What is true of 
a thing is true of it, surely, under any name. (1995: 90-1) 

Quine's insistence on extensionality is a very long-running theme of his thought, going 
back as far as the clarification of Principia Mathematica (Whitehead and Russell 
1910-13) in his doctoral dissertation. (See the first paragraph of this essay; also Quine 
1991: 265-6.) 

The requirement of extensionality has been controversial because there are large 
areas of discourse which, at least if taken at face value, are not extensional. This for 
Quine is reason enough not to take such discourse at face value. Instead, he thinks, we 
should either try to reformulate it so that it becomes extensional or else exclude it from 
the more scientific and respectable parts of our knowledge, those parts which we take 
as really telling us about the objective world. We shall discuss three examples; our dis- 
cussion of each will be very brief, although the second and third of them are issues 
which have generated much controversy, and on which Quine has, largely for that 
reason, written extensively. 

The first sort of prima-facie violation of extensionality is almost trivial; indeed it 
might be said that the idea that we really have such a violation here is simply a mistake. 
This is the case of quotation, which is worth examining because it functions as some- 
thing of a paradigm for Quine. It is true to say: " 'Quine' has one syllable"; it is false to 
say " 'The author of Word and Object' has one syllable"; yet the one sentence might seem 
to be obtained from the other by replacing a singular term with another singular term 
designating the same object, since Quine, of course, is the author of Word and Object. 
Here the solution to the apparent puzzle is easy. We should not construe the subjects of 
the sentences as referring to a person, but rather to the words, the expressions them- 
selves. It is of the word "Quine," not of the philosopher Quine, that we say it has one 
syllable. And then of course the substitution no longer replaces a name by another 
name for the same object, since the two expressions are not the same. 

The second sort of example is far less easy to dismiss. The case of indirect discourse, 
where one reports the speech or the thoughts of another, also gives rise to prima-facie 
cases of non-extensionality. Othello (supposing him for the moment to be a real person) 
has the false belief that Desdemona loves Cassio, but does not (presumably) have the 
equally false belief that the moon is made of green cheese. Clearly, in statements of the 
form "A believes thatp" it is not only the truth-value of p that is relevant to the truth- 
value of the whole. Similarly in the case of singular terms. To use an example of 



Quine's, Tom may believe that Cicero denounced Catiline without believing that Tully 
denounced Catiline, for he may not know that Cicero is Tully. 

One response to such cases is to say that what is believed or disbelieved is not a fact 
about a person (or other object) but rather a proposition. (For this reason, philosophers 
often speak of such cases as statements of "propositional attitudes.") Then it is claimed 
that we have two distinct propositions: first, that Cicero denounced Catiline, and, second, 
that Tully denounced Catiline. So construed, belief-contexts become extensional, for we 
can no longer obtain a falsehood from a truth by substituting a co-designative expression 
(since this now means an expression referring to the same proposition) . Quine rejects this 
idea, chiefly on the grounds that no clear and precise identity-conditions have been given 
for propositions, which should therefore not be accepted for scientific and philosophical 
purposes. This is of a piece with his having some degree of skepticism about meaning, 
for one can think of a proposition as being the meaning of a declarative sentence. (In that 
case the question of the identity-conditions of propositions is the same as the issue of 
synonymy for declarative sentences.) 

Quine's treatment of indirect discourse relies not on the meanings of sentences but 
simply on the sentences themselves. Formally, he assimilates them to cases of quota- 
tion, by construing belief (and doubt, and hope, and so on) as attitudes towards sen- 
tences. Perhaps there is something odd or counterintuitive about speaking of believing 
or not believing a sentence, but this kind of oddity is something that Quine is, as we 
have seen, fully prepared to accept in the interests of clarity (as he conceives it). And if 
we do talk this way, then there is nothing puzzling in thinking that Tom, to revert to 
our example, may believe the one sentence without believing the other, for they are dis- 
tinct objects. 

Two points call for comment. First, Quine was for a period convinced that there are 
two kinds of belief. There is the ordinary kind, variously construed as an attitude 
towards a sentence or a proposition. But then there is also, he thought, another kind, 
de re belief, which is belief genuinely about an object, so that the way in which the object 
is described is irrelevant to the truth of the belief-attribution. He spent much effort 
trying to make sense of cases which appear to be of this sort, but subsequently came 
to think (quite correctly, in my view) that the appearance of two distinct kinds of belief 
is mistaken, and that the phenomena that he was attempting to understand are in fact 
not really cases of a different kind of belief. Second, it would be a mistake to think of 
the difficulties of making clear sense of belief contexts as merely formal. It is Quine's 
view, after all, that we want an extensional language not for its own sake alone, but 
because lack of extensionality is a sign of lack of clarity. His re-construal of belief as 
an attitude towards sentences puts the emphasis in what he thinks, for independent 
reasons, in the right place. If we are construing evidence as austerely and strictly as 
possible, the evidence for a belief-ascription is simply a report of what the person con- 
cerned said - the very words uttered - and did. In reports of these sorts there is no vio- 
lation of extensionality. Beyond that, belief-ascription relies on empathetic projection: 
the ascriber imagines what it would be like to be in the believer's situation. Here, it may 
be thought, we are out beyond the realm of hard fact. 

A third area of discourse in which extensionality fails, at least on Quine's view, is 
that concerning modality. Suppose it said that nine is necessarily greater than five (the 
example, again, is Quine's). Replacing "nine" by a co-designative expression we can 



obtain: the number of the planets is necessarily greater than five. Quine, however, 
sees no hope of making sense of any notion of necessity according to which this is 
true (cf. MARCUS). He distinguishes a notion of necessity of this sort from one which 
attaches to sentences (and hence can be assimilated to the case of quotation) or treats 
it as a statement operator. In these cases the necessity is supposed to hold of the 
sentence, not of the object, so there is no risk of the sort of violation of extensionality 
indicated above. The general issue of modality, however, is not a crucial one from 
a Quinean point of view, for he sees no need to accept any notions of necessity or 
possibility. He holds that a reconstruction of our knowledge, or of those aspects of it 
which we want to take with full seriousness, as telling us about the objective world, 
require no such notions. 

We have briefly discussed three cases of prima-facie non-extensionality It is worth 
emphasizing, however, that these three do not exhaust the matter. Counterfactual con- 
ditionals, statements of causality, and, if taken at face value, statements of time and 
tense, are among other such cases. 

Ontology and its relativity 

A crucial part of the task of getting clear about what our theories commit us to, on 
Quine's account, is gauging their ontological commitments. By the ontological commit- 
ments of a theory Quine means what entities that theory says there are in the world. 
In Quine's view, the way to settle this is by seeing what objects must be in the range of 
the theory's variables if it is to be true. Quine's emphasis on ontology, and on the range 
of variables as the measure of ontological commitment, is perhaps in part to be 
explained by his early work in set-theory, and his abiding interest in that subject. In that 
context, the range of the variables of the theory is a natural measure of its strength, 
and the threat of paradox makes this matter of vital concern. (The importance of this 
point was emphasized to me by Stephen Menn, to whom I am grateful.) 

Quine takes ontology to be a product of self-conscious scientific and philosophical 
reflection. He does not see the philosopher's ontological task as that of capturing in per- 
spicuous form the ontology implicit in ordinary thought and discourse, for he insists 
that there is no such "ordinary ontology": 

a fenced ontology is just not implicit in ordinary language. The idea of a boundary between 
being and non-being is a philosophical idea, an idea of technical science in a broad sense. 
Scientists and philosophers seek a comprehensive system of the world, and one that is 
oriented to reference even more squarely and utterly than ordinary language. Ontological 
concern is not a correction of a lay thought and practice; it is foreign to the lay culture, 
though an outgrowth of it. (1981: 9) 

When we are concerned with ontological questions, or metaphysical questions more 
generally, we cannot simply examine our beliefs in the terms in which we are at first 
prone to express them. Our beliefs must, rather, be cast into a standard notation, which 
will let their presuppositions shine forth. Ontology, as Quine interprets it, thus presup- 
poses regimentation: it is only insofar as we conceive of our knowledge as cast in 
regimented notation that it makes sense to raise ontological questions. 



The artificiality of ontology, in Ouine's view, is an important point. He is often criti- 
cized for distorting ordinary thought, or doing violence to our supposed "intuitions." 
On his view these criticisms miss their mark entirely. More positively he takes part of 
the task of ontology to be that of showing just what our theories really commit us to. 
Here he has an interest in ontological economy. In many cases our theories seem to 
commit us to accepting entities of a certain kind, but artful re-construal of the theo- 
ries shows that in fact we need not presuppose entities of the given sort. Here artifi- 
ciality is inevitable. 

In this spirit, Quine takes the reduction of numbers to sets, and of ordered pairs 
to sets, to be clear philosophical achievements. (Section 53 of Quine 1960 is entitled 
"The Ordered Pair as Philosophical Paradigm.") In each case the reduction shows 
how our discourse could be rephrased so as to avoid commitment to a kind of entity 
to which it appears to be committed. The rephrasing is not meant as a practical 
substitute for normal arithmetical or set-theoretic language. Nor is it claimed that 
the paraphrased version gets at the "real meaning" or hidden structure of the original. 
It is simply that the reduction shows us how we could do everything that we need, for 
scientific purposes, without presupposing that there are ordered pairs (or numbers), 
and hence that we need not take ourselves to be committed to the existence of 
such things. Nor is it only towards such technical subjects as arithmetic and set-theory 
that Quine takes this attitude. On the contrary; he takes the same view everywhere. 
He holds, for example, that we can eliminate minds in terms of bodies, just as we 
can eliminate numbers in terms of sets. Instead of speaking of a person's mind at 
a given moment, we could instead speak of his or her body at that moment. Mentalistic 
predicates, such as "is thinking of Vienna" persist, but no exclusively mental objects. 
The maneuver is trivial, but not, in Ouine's view, any the worse for that; he takes 
it as showing that we have no ontological commitment to minds, over and above 

Quine thus stresses the artificiality of ontology, and the use of paraphrase or 
re-construal to eliminate apparent ontological commitments. His tactics here presup- 
pose something implicit in our earlier discussion of knowledge, especially of obser- 
vation sentences. In his view language is, of course, referential - it is about things. We 
refer to people and other objects in almost everything we say. But the relation 
of reference is not our fundamental cognitive relation to the world. The fundamental 
relation, the way that language gets to be about the world at all, is the relation 
of observation sentences to patterns of stimulation. (In particular, the relation of 
a given observation sentence to the patterns of stimulation on the basis of which 
speakers of the language are inclined to accept or reject the given sentence.) This 
is not a relation of reference, for observation sentences are not about patterns 
of stimulation. Reference is a derivative relation. Indeed the very notion of an 
object, that to which we refer, is in Ouine's view derivative - though it comes so 
naturally to us that it seems inevitable. If there were a language consisting only of 
observation sentences, there would be no reason to attribute any reference to that 
language. The vastly greater complexity of our language requires inferential 
links between sentences, links which we can grasp only by attributing structure to 
the sentences. Seeing sentences as divided into terms, some of which refer, is part 
of this process: 



Reference and ontology recede thus to the status of mere auxiliaries. True sentences, 
observational and theoretical, are the alpha and the omega of the scientific enterprise. 
They are related by structure, and objects figure as mere nodes of the structure. 
(Quine 1990:31) 

The derivative character of ontology manifests itself most dramatically in the 
Quinean doctrine known as ontological relativity, or the inscrutability of reference. 
Quine's claim is that it would be possible to carry out a large-scale re-construal of our 
knowledge, replacing every referring term by another. We could, for example, replace 
every term referring to a physical object by a term referring to its space-time comple- 
ment, that is, to the whole of space and time other than the given object. Along with 
that we would re-construe our predicates, so that a predicate true of a given object 
would now be taken as true of its space-time complement. These re-construals would 
cancel out, leaving the truth-value of each sentence the same as before. The result 
would be a system of knowledge exactly like our own in its structure, including its rela- 
tion to evidence, but in which each term refers to objects different from those to which 
its unreconstructed analogue refers. Nothing, Quine claims, prevents such a re- 
construal. The passage quoted in the previous paragraph continues: "What particular 
objects there may be is indifferent to the truth of observation sentences, indifferent to 
the support they lend to theoretical sentences, indifferent to the success of the theory 
in its predictions" (Quine 1990: 31). 

Quine claims that this sort of wholesale re-construal of our knowledge is possible, 
that it would not be inconsistent with any part of it. True, f can and would stoutly main- 
tain that my word "rabbit," say refers to rabbits and not to their space-time comple- 
ments. But all my uses of the word, including those uses in which I insist that it is rabbits 
that I am talking about, are open to re-construal. From the possibility of this sort of re- 
construal, Quine sometimes infers that there is no fact of the matter about reference - 
no fact of the matter as to whether my word "rabbit," say really refers to rabbits, or 
rather to the space-time complement of all the rabbits. The thesis of ontological rela- 
tivity, however, is perhaps best thought of as a rejection of this idea of "really refers," 
insofar as it outruns the ordinary idea of reference. Taking our own language for 
granted we can say to what the words of another language (or, trivially, of that same 
language) refer; while we stay within that language we use it to refer with no more than 
the usual difficulties or ambiguities. Ontological relativity is, Quine says, "unproblem- 
atic but trivial" (Hahn and Schilpp 1998: 460). Only when we translate, or map our 
language onto itself, does the ontological relativity emerge. And here it should perhaps 
be seen as a reminder of the derivative status of reference. 

Quine has come to give increasing emphasis to the doctrine of ontological relativity. 
Should we see it as conflicting with, or undermining, his insistence on realism? Since 
it leaves truth-values unaffected, it does not affect his claims to be a realist about truth. 
That latter claim, as we saw, was also one that might be doubted. Quine's claim there 
is perhaps that he is as much of a realist as it makes sense to be: that there simply is no 
coherent sense of realism stronger than his. Similarly here, perhaps. Quine claims that 
ontological relativity is beyond doubt; it "admits of trivial proof," he says (Hahn and 
Schilpp 1998: 728). So there is in his view no chance of defending a version of realism 
that denies it. Ontological relativity might thus be taken as showing us what realism 



can come to. Ontology simply is derivative upon truth, and given any system of truths 
there simply will be more than one way of construing its ontology. If this undermines 
our previous conception of realism then, from Ouine's point of view, so much the worse 
for that conception. 


I shall not attempt to summarize what is already a very compressed treatment of 
Ouine's thought. It is perhaps worth saying, however, that I have ignored, or treated 
very briefly, a number of issues that have occupied considerable space, both in Ouine's 
own writings and in the works of commentators and critics. Notable examples here 
include the indeterminacy of translation, Ouine's views of indirect discourse and of 
modality, the question of his physicalism, and what it amounts to, and his substantive 
views on ontology. My decisions as to how to use the limited space available here are of 
course based on my view of what is most important in Ouine's thought. But the reader 
should perhaps know that others might have made these decisions rather differently 1 


1 For their comments on earlier drafts of this essay I am indebted to Bill Hart, Peter Hacker, 
Gary Kemp, and Al Martinich. 


Works by Quine 

1936: "Truth by Convention," in Philosophical Essays for A. N. Whitehead, ed. O. H. Lee, New York: 

Longmans. (Reprinted in Quine, The Ways of Paradox, rev. edn., Cambridge, MA: Harvard 

University Press, 19 76.) 
1960: Word and Object, Cambridge, MA: MIT Press. 
1961: From a Logical Point of View, 2nd, rev., edn., Cambridge, MA: Harvard University Press. 

(First published 1953.) 
1963: "Carnap and Logical Truth," in Paul A. Schilpp (ed.) The Philosophy of Rudolf Carnap, La 

Salle, IL: Open Court. (Reprinted in Quine, The Ways of Paradox, 1976.) 
1969: Ontological Relativity and Other Essays, New York: Columbia University Press. 
1974: Roots of Reference, La Salle, IL: Open Court. 
1981: Theories and Things, Cambridge, MA: Harvard University Press. 
1987: Quiddities: An Intermittently Philosophical Dictionary, Cambridge, MA: Harvard University 

1990: Pursuit of Truth, Cambridge, MA: Harvard University Press. 
1991: "Two Dogmas in Retrospect," Canadian Journal of Philosophy 21, pp. 265-74. 
1995: From Stimulus to Science, Cambridge, MA: Harvard University Press. 
2000: "Response to O'Grady," The Proceedings of the World Congress of Philosophy, Boston, 1998, 

Bowling Green, OH: Philosophy Documentation Center. 

Works by other authors 

Carnap, Rudolf (1934) Die logische Syntax der Sprache; Vienna: Springer Verlag. (Published, with 
additions, as The Logical Syntax of Language, trans. Amethe Smeaton, London: Routledge and 
KeganPaul, 193 7.) 



Creath, Richard (ed.) (1990) Dear Carnap, Dear Van, Berkeley and Los Angeles: University of 

California Press. 
Gibson, Roger (1982) The Philosophy of W.V. Quine, Tampa: University Presses of South Florida. 
Gregory, Paul (1999) "Ouine's Natural Epistemology," Ph.D. dissertation, University of Illinois. 
Hahn, Edwin and Schilpp, P. A. (eds.) (1998) The Philosophy of W. V Quine, expanded edition, 

Chicago and La Salle, IL: Open Court Press. (Previously published 1986.) 
Hylton, Peter (1990-1) "Translation, Meaning, and Self-Knowledge," Proceedings of the 

Aristotelian Society 9 1 . 
Lee, Harold N. (1998) "Discourse and Event: The Logician and Reality," in Hahn and Schilpp, 

pp. 295-315. 
Smart, J. J. C. (1969) "Ouine's Philosophy of Science," in Words and Objections: Essays on the Work 

of W.V. Quine, ed. Donald Davidson and Jaakko Hintikka, Dordrecht: Reidel, pp. 3-13. 
Whitehead, Alfred North and Russell, Bertrand (1910-13) Principia Mathematica, 3 vols, 

Cambridge: Cambridge University Press. 2nd edn. 192 7. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

A.J. Ayer (1910-1989) 


Language, Truth and Logic 

General character of the book 

A. J. Ayer rose to early philosophical fame with the publication in 1936, when he was 
25 years old, of what remained his most famous, or infamous, book, Language, Truth 
and Logic. The work is his own version of the logical positivism characteristic of the 
Vienna Circle (whose meetings he had attended for three months in 1932-3), his 
outlook being closest to that of their leader, Moritz Schlick. The book is also strongly 
influenced by the British empiricist tradition, in particular by Hume and Russell. It was 
something of a bombshell to British philosophers and became for them the paradigm 
statement of logical positivism, threatening the outlook of some, providing an exciting 
intellectual liberation for others. 

The book opens with the striking statement: 

The traditional disputes of philosophers are, for the most part, as unwarranted as they are 
unfruitful. The surest way to end them is to establish beyond question what should be the 
purpose and method of philosophical enquiry. (1946: 33) 

So far as philosophy goes, Ayer's concern is to show the meaninglessness of meta- 
physical theories about a reality beyond the empirical. More generally, he also claims 
to show that religious statements, as usually now intended, are meaningless, as also 
are statements of fundamental ethical principle (except as mere expressions of 
emotion). 1 

To establish the meaninglessness of all such statements Ayer puts forward the veri- 
fication principle. According to this a statement 

is factually significant to any given person, if, and only if, he knows how to verify the 
proposition which it purports to express - that is, if he knows what observations would 
lead him, under certain conditions, to accept the proposition as being true, or reject 
it as being false. If, on the other hand, the putative proposition is of such a character 
that the assumption of its truth, or falsehood, is consistent with any assumption whatso- 
ever concerning the nature of his future experience, then, as far as he is concerned, 



it is, if not a tautology, a mere pseudo-proposition. The sentence expressing it may be 
emotionally significant to him; but it is not literally significant. (1946: 35) 

Thus a meaningful statement must either be empirically verifiable, or be a tautology, 
that is, analytic or true by definition. 

The passage just quoted is supposed to be a "somewhat vague" formulation of a 
principle that Ayer proceeds to express more precisely. But actually, since it turned 
out difficult to find a satisfactory precise formulation, it remains as good a formulation 
as any. 

Since there are two types of meaningful statement for Ayer, the empirically verifi- 
able and the analytic, his account of each of these will be considered in turn. 

Empirical statements 

Ayer distinguishes between strong and weak verification. A strongly verifiable proposi- 
tion is one which could be conclusively established by sense experience, a weakly 
verifiable proposition is one which could be made probable by sense experience, ft is 
too much to demand strong verifiability of a meaningful factual statement (it is 
doubtful indeed if any proposition is strongly verifiable) and so some form of weak 
verifiability is the appropriate criterion. Ayer tries to give an exact formulation of this 
as follows. 

Let us call a proposition which records an actual or possible observation an experiential 
proposition. Then we may say that it is the mark of a genuine factual proposition, not that 
it should be equivalent to an experiential proposition, or any finite number of experiential 
propositions, but simply that some experiential proposition can be deduced from it in 
conjunction with certain other premises without being deducible from these other 
premises alone. (1946: 38-9) 

The general idea, here, is clear enough. A meaningful empirical statement must be 
a genuine aid to the anticipation of the experiences we can expect to have under various 
circumstances (identified in terms of the other experiences then available), though it 
need not tell us what experiences to expect all on its own. To illustrate his point Ayer 
gives an example of two questions that might be raised about a painting. (1) Was it 
painted by Goya? (2) fs the painting a set of ideas in God's mind? People may disagree 
in their answers to each of these questions, but in the first case they know what kind 
of empirical evidence would support their claim against that of their opponents, in the 
latter they do not (1946: 40). 

Later in the work, especially in chapter VIII, "Solutions of Outstanding Philosophical 
Disputes," Ayer shows, or claims to show, how metaphysical questions are all mean- 
ingless in much the same manner, unless they are understood, as is often appropriate, 
as misleading ways of discussing how propositions of a certain type are to be analyzed. 
(See the section below, "What is the task of philosophy?") 

Unfortunately Ayer later discovered that the technical formulation of this revelation 
of the meaninglessness of metaphysical questions was unsatisfactory. This is because 
any statement whatever, call it "P," can meet the condition simply in virtue of the fact 
that its conjunction with "ff P, then 0" (where "0" is a experiential statement) 2 entails 



"0," as "If P then 0" does not alone. Thus "God is annoyed," which Ayer would hate to 
find meaningful, entails an observation statement "You will shortly hear thunder" when 
conjoined with "If God is annoyed with what you said, you will shortly hear thunder." In 
his 1946 introduction to the second edition Ayer offered a more complicated formula- 
tion, which, however, he had later to concede, fell foul of technical criticisms from Alonzo 
Church and C. G. Hempel. (See Church 1949 and Hempel 1959.) 

If one is not too infatuated with semi-formalization, however, one can, surely, 
say clearly enough what Ayer was getting at, whether one accepts it or not. Surely the 
real point of the verification principle, so far as factual (non-analytic) statements go, 
was this. Such a statement is meaningful to a particular individual if and only if 
it is possible for either it or its negation to be a practical aid to him in forming correct 
expectations about what he is liable to experience in the future. If there is no such 
possibility then it is factually meaningless, however much he may suppose himself to 
understand it. 3 

If there is a problem, here, it is about what "possible" means, but perhaps it is 
sufficient that the individual does not utterly rule out its occurring. It is to be noted, 
however, in this connection, that Ayer is anxious to distinguish practical verifiability 
from verifiability in principle . Thus, in an intriguingly dated example, "There are moun- 
tains on the other side of the moon" was said to be unverifiable in practice but verifi- 
able in principle (an example taken from Moritz Schlick). 

An important question is whether the verification principle is intended not only to 
tell us whether a statement is meaningful or not, but also to tell us what its meaning 
is. In effect, rather than in actual formulation, Ayer treats it as doing so and surely this 
is correct. For if a factually meaningful statement must be a possible aid to knowing 
what experiences to expect under various circumstances then its meaning must lie in 
the totality of such aid as it is capable of giving. If there is some residue of purported 
further meaning it would seem that this could be creamed off as an unverifiable state- 
ment included within it. 

That the verification principle is intended to exhibit the meaning of factual proposi- 
tions is plain from Ayer's deductions from it concerning the analysis of a whole range 
of ordinary statements of fact. Thus the verification principle is said to make inevitable 
a phenomenalist analysis of statements about material objects, since it is only "by the 
occurrence of certain sense-contents that the existence of any material thing can ever 
be in the least verified" (1946: 53). It is no good some objector saying that the existence 
of a physical object is not merely a fact about what sense-contents are available to us, 
though it is by this that it is verified, for ultimately all that can be verified by facts about 
sense-contents are facts about sense-contents and the probable truth of what can be 
inferred from such facts inductively. (See the following section). Ayer, however, like 
Quine later, eschews talk of meanings as entities, substituting for talk of meanings talk 
of synonymy (1946: 68) (see QUINE). 

Analytic or a priori statements 

The other sort of meaningful statements for Ayer were analytic statements. All 
genuinely necessary or a priori statements are of this type; thus anything like the 
synthetic a priori of Kant and others is rejected. 



One initial point worth remarking is that, while Ayer is clear that, if a statement is 
empirically meaningful then so are its contraries and contradictory, his assertion that 
the only two types of meaningful statements are empirical hypotheses and analytic 
propositions, taken strictly implies that this is not so in the case of the latter. If so, while 
"5 + 3 = 8" is meaningful, the proposition "5 + 3 = 9" is not false, but meaningless. 
How far this is intended is unclear, since Ayer does, in fact, talk of false mathematical 
statements. (See 1946: 86.) 

Ayer's discussion of analytic propositions (in 1946: ch. IV) starts out from the 
problem which a priori truth is supposed to pose for empiricism (of which his logical 
positivism is avowedly a species). For empiricism can countenance no claim to knowl- 
edge that is not based upon sense experience, and even then what is called "knowledge" 
is always probable hypothesis rather than absolute certainty. 

Where the empiricist does encounter difficulty is in connection with the truths of 
formal logic and mathematics. For whereas a scientific generalisation is readily admitted 
to be fallible, the truths of mathematics and logic appear to everyone to be necessary and 
certain. But if empiricism is correct no proposition which has a factual content can be 
necessary or certain. Accordingly the empiricist must deal with the truths of logic and 
mathematics in one of the two following ways: he must say either that they are not 
necessary truths, in which case he must account for the universal conviction that they 
are; or he must say that they have no factual content, and then he must explain how a 
proposition which is empty of all factual content can be true and useful and surprising. 
(1946: 72-3) 

Having dismissed the first alternative, that of J. S. Mill, according to which, for 
example 2 x 3 = 6 is simply so well confirmed a statement of fact that we (wrongly) 
think that it could not have been otherwise, Ayer opts for the view that all a priori and 
necessary (these are identified) so-called truths are really analytic. 

Rejecting Kant's account of analyticity for various reasons, Ayer formulates his own 
account. This, however, is somewhat shifting. The most definitive formulation would 
seem to be this: 

a proposition is analytic when its validity depends solely on the definitions of the symbols 
it contains, and synthetic when its validity is determined by the facts of experience. Thus, 
the proposition "There are ants which have established a system of slavery" is a synthetic 
proposition. For we cannot tell whether it is true or false merely by considering the defi- 
nitions of the symbols which constitute it. We have to resort to actual observation of the 
behaviour of ants. On the other hand, the proposition "Either some ants are parasitic or 
none are" is an analytic proposition. For one need not resort to observation to discover that 
there either are or are not ants which are parasitic. If one knows what is the function of 
the words "either," "or," and "not," then one can see that any proposition of the form 
"Eitherp is true orp is not true" is valid, independently of experience. Accordingly, all such 
propositions are analytic. . . . However, when ... we say that analytic propositions are 
devoid of factual content, and consequently that they say nothing, we are not suggesting 
that they are senseless in the way that metaphysical utterances are senseless. For although 
they give us no information about any empirical situation, they do enlighten us by illus- 
trating the way in which we use symbols. (1946: 78-9) 



There are several difficulties with propositions of the type he classifies as both 
"analytic" and "a priori." On his account, are these propositions not, in effect, 
statements about how certain symbols are normally used? But if so, they seem to be 
empirical, since it is an empirical fact that we use words as we do. Ayer tackles this 
question in his introduction to the second edition. (See 1946: 16-18.) His reply is that, 
although they are simply the consequences of sticking to a certain consistent use 
of certain symbols, they do not so much state as presuppose such rules of language. 
And this, thinks Ayer, explains why they can be surprising. For there are (doubtless 
infinitely) many consequences of this sort which it requires considerable intellectual 
power to grasp. 

What is troubling about this answer is that while the simpler statements that Ayer 
calls "analytic" may be thought of as little more than reminders of how we optionally 
use certain symbols, there are innumerable consequences of such use which follow 
therefrom in a manner that is not similarly optional. Compare the rules of chess. There 
is no proper answer, other than a presently irrelevant historical one, as to why the 
pieces may be moved just as they may, but from this set of optional rules untold con- 
sequences follow as to how the game can best be played to win. Or to take a case more 
to the point, even if the proposition that 7 is 6 + 1 is simply a reminder of the meaning 
of "7," the proposition that 7 is a prime number is not a reminder of how we use the 
symbol but a necessary consequence thereof. It is of this latter sort of necessity that 
Ayer gives no satisfactory account. Or so at least it seems to some of us. 4 

Be that as it may, the general idea is clear, namely that all necessary or a priori truths 
are really the consequences of an optional use of language and tell us nothing about 
anything non-linguistic. Nor do they exactly say anything about language, they simply 
helpfully reflect back to us the character of the language in which our knowledge or 
beliefs are expressed. 

Perhaps Ayer's position is, in effect, that the function of analytic statements is not 
strictly to say anything but to serve as a kind of verbal drill whereby we reinforce and 
improve our command of the rules and accepted transformations which give verbal 
expressions their meaning.' 

What is truth? 

So much for language and logic. What of truth? One might suppose that for Ayer truth 
would consist in being a reliable predictor of sense experience in the case of empirical 
propositions, and following from linguistic rules in the case of analytic propositions. 
Actually, in chapter V Ayer puts forward what has been called the redundancy theory 
of truth. According to this, the correct account of what "true" means is that it simply 
emphasizes the assertion of the proposition said to be true. Thus to say that the propo- 
sition that dogs bark is true is simply to say that dogs bark. Similarly to say that a propo- 
sition is false is simply to assert its negation. To say that it is false that cats bark is simply 
to say that cats don't bark. 

But how are we to explain occurrences of "true" where the proposition said to be 
true is not formulated, for example, "Everything he said in the lecture is true"? Ayer's 
implied answer is that this means "(p)(he said that p implies p)." There is a problem with 
this answer, into which we shall not enter, in that the variable "p" occurs firstly as a 



name variable and secondly as a propositional variable; that is, the first occurrence of 
"p" stands in for the name of a proposition, and its second occurrence for the actual 
formulation of a proposition. 

In the light of this Ayer says that philosophical attempts to answer the question 
"What is truth?" are largely misconceived. When they have a definite meaning, the 
question asked is really, "How are propositions validated?" The meaning of "validated" 
is not made clear, but the question seems to mean "What are the criteria we properly 
use in deciding whether to affirm or deny them?" (Ayer often uses "validate," as it seems 
to me, to avoid talk of judging to be true.) The answer is that analytic propositions are 
validated by their being consequences of the way we use words, while empirical propo- 
sitions are validated by the fact that they have been found successful as a way of pre- 
dicting what we will experience under various circumstances, and are thereby taken as 
likely to be similarly successful in the future. (See 1946: 99.) If we raise "the problem 
of induction" associated especially with Hume, as to what right we have to take past 
experience as a guide to the future, Ayer's answer is, roughly, that to be guided in this 
way is just what we presently mean by being "rational." 

What is the task of philosophy? 

If metaphysics is nonsense, and to be abandoned, is there any type of philosophy that 
is more intellectually respectable? Ayer's positive answer is that the sort of philosophy 
that is a worthwhile activity is (conceptual) analysis. (See 1946: chs II and III.) And in 
fact this was what genuinely great philosophers have always been mainly engaged in. 
Often they have put their questions in the form of "What is X?" e.g. "What is matter?", 
"What is time?", "What is the self?", and there is nothing wrong in this mode of expres- 
sion, if it is properly understood. Thus understood, these questions are really requests 
for definitions of some of the very general expressions in our language which either 
puzzle us or lead to metaphysical nonsense. 

The traditional type of definition professes to explicate the meaning of a word by 
offering some other more complex verbal expression which is its equivalent. Ayer's 
trivial example is An oculist is an eye doctor," which tells us that "oculist" and "eye 
doctor" mean the same (1946: 60). Definitions of this type are for the most part of 
limited use to philosophy, which is concerned with a more fundamental clarification of 
what both expressions are meant to stand for. Instead the philosopher requires so-called 
"definitions in use." (See pp. 60-3.) A definition of this sort is an instruction for trans- 
lating statements about X or Xs into equivalent statements which have no word or 
expression referring (grammatically considered) to X or Xs. These are useful when Xs 
strike a philosopher as somehow not belonging to the bedrock of reality. 

Thus a philosopher might try to answer the question "What is a nation?" by showing 
how specimen statements about nations can be translated into statements about people. 
For example, he might seek a way of translating "Britain and Germany were at war 
from 1939 to 1945" into a complicated statement about how people whose homes were 
on one part of the earth's surface behaved towards, and were affected by, people whose 
homes were on another part of the earth's surface. If such a translation of statements 
about nations into statements about people and land is possible, then a nation may 
be described as a "logical construction" out of people and land, though it should be 



realized that this is not a statement about how two sorts of thing are related but between 
linguistic expressions, one of which is supposedly puzzling in a way which the other 
is not. 

Even in such rather obvious cases as that of nations such definitions are usually 
gestured towards rather than actually formulated. The gesture may be sufficient, 
however, to show that what the statement about the nations says in a simple, but mis- 
leading (because it suggests that nations are something over and above persons and 
land) way, is something which in principle should be sayable in a way that gives us no 
excuse for being thus misled. As for the difficulty in finding quite satisfactory actual 
translations, this, it may be suggested, may be because they would have to be impos- 
sibly complicated, or because they would have to be precise about details that the state- 
ments being explained leave vague. So although war is certainly a matter of people 
doing things one cannot be precise about just what people must do to be at war. 
(Cf. 1954: 141-3.) 

So the task of philosophy is to point towards definitions in use of expressions 
that are liable to puzzle us or to suggest that there are things over and above those we 
actually encounter empirically. 

Phenomenalism regarding physical objects 

Central to Ayer's type of logical positivism is a phenomenalist view of physical or ma- 
terial objects. According to this, every statement about the physical world is, in princi- 
ple, translatable into a proposition to the effect that under such and such conditions 
such and such sensations will or would occur. Thus the proposition that a physical 
thing exists always means "that, if certain conditions were fulfilled, certain sense- 
contents . . . would be experienced" (1946: 141). 6 

"Experienced by whom?" one may well ask. The answer would seem to be "by whom- 
soever it is who is affirming the proposition"; in short, when you affirm it, it tells 
you what sensations you should expect under such and such circumstances, while 
when f affirm it, it tells me what sensations I should expect under such and such 
circumstances. This suggests that each of us gives our own private meaning to the 
proposition. And in fact the doctrine of the book, without perhaps the author being 
fully aware of it, is that what counts as the same factual statement has a different 
meaning for each person, since for each of them the information it provides, if 
it is true, concerns just their own actual and possible experience. This is disguised, 
somewhat, by the fact that persons are themselves supposed to be logical construc- 
tions out of sensations, though not the same sort of logical construction as physical 
objects are. 

For a person, so far as his conscious mind goes, is, according to Ayer, a logical con- 
struction out of those sense-contents that occur in the same sense-fields as do the 
organic sense impressions of that body. Among the consequences that follow from this 
is that an individual's survival of bodily death is a meaningless idea, insofar as there 
can be no organic sensations of his body thereafter. Here Ayer differed from the leader 
of the Vienna Circle, Moritz Schlick, who thought it perfectly meaningful to suppose 
that I might verify my own death by having sense impressions as of seeing my funeral 
from a point of view unoccupied by a human body. (See Schlick 1949: 159-60.) 



Propositions about other minds 

This leads naturally to the account of our knowledge of other minds presented in 
Language, Truth and Logic, chapter VII. It is likely to be charged, he says in effect, that if 
I accept the doctrine of that book then I am committed to solipsism, to the view that 
only I exist, as a conscious individual with my own sense experiences. For the existence 
of the sense-experiences of other people appears to be, in principle, something that 
I cannot verify, since my sense-impressions can only be associated with the organic 
sensations of my own body, not with those of another person. But Ayer rejects this 
conclusion, contending that each of us must define the existence of other persons, 
including the sensations that go with the organic sensations of their body, in terms of 
the behavior on the basis of which I would ordinarily conclude that they were conscious 
and had sensations. 

Thus he held, at that stage, that what "I" (whoever I am) mean when I speak of 
my own sensations is that such and such sensations actually occur together with 
the organic sensations of my body, but that when I speak of the sensations of another 
person I mean that they are behaving, or are disposed to behave, in such and such a 
way (this being a physical fact ultimately consisting in facts about my sensations as of 
perceiving their bodies move and make noises etc. of such and such a sort). 

This view about what we mean by speaking about the sensations of others, reveals, 
as Ayer later himself insisted, a peculiar double-take on the whole business of what we 
mean by what we say (see 1956: 245-7). The official view of Language, Truth and Logic 
is that everyone means by everything they say something about their own sensations. 
But this is a view delivered as true of every speaker, by Ayer, qua philosopher, who, 
thereby, is clearly supposing that other people have sensations in the same sense as he 
does (i.e. in a way not analyzable behavioristically) though it is part of the theory that 
this realist conception of the sensations of others is, for each of us (and that must 
include Ayer himself) meaningless (see 1946: 141). 

Propositions about the past 

As strange, or stranger, than this view of the meaning of assertions about the experi- 
ences of others is the view that propositions about the past can only be meaningful 
(because otherwise unverifiable in principle) if they are equivalent to predictions about 
the kind of so-called historical evidence that would support them. An oddity (we 
may remark) of this view is that, while empirical knowledge is said to consist in 
predictions about future experience on the basis of past experience, the fact that the 
past experiences occurred is itself a prediction of the same essential kind. 

The paradoxical character of these conclusions is among the factors which led Ayer 
away from the precise positions of Language, Truth and Logic, though he struggled to 
remain true to at least the general spirit of the verification principle. 

Critique of ethics and theology 

One position, however, which Ayer never abandoned was the emotive theory of ethics 
advanced in chapter VI (see STEVENSON). According to this ethical concepts are pseudo- 



concepts, that is, they lack factual meaning. To say that behavior of a certain sort is 
wrong is not to state anything about it which can be true or false. It is simply to express 
a feeling. True, there may be ethical statements which include a factual element. If I 
say "He did wrong to kill the cat," then, inasmuch as this says that he killed the cat it 
is meaningful, and true or false. But calling it wrong adds nothing cognitively mean- 
ingful. And if I make a statement of pure ethical principle, such as "Suicide is wrong" 
then this is just as though I held up my hand in horror at the idea of people commit- 
ting suicide. This has been called (not by Ayer) the "boo-hurrah" theory of ethics. He 
gave it a milder statement in a later essay (1954: essay 10) pointing out that he was 
not making a negative value judgment about moral thinking, and, in fact, Ayer was 
personally and publicly committed to strong liberal principles. 

As for religious propositions, statements like "God exists" or "God loves us," as most 
people now think that they understand them, they are meaningless. There is no obser- 
vational test that could be used to determine their truth or falsehood. God is not, as the 
mountains on the other side of the moon were then, something whose existence is in 
fact unverifiable, but in principle verifiable. It is worth noting that even if Ayer revised 
his view that life after death was meaningless he could still put up a good case for saying 
that the existence of God would remain so, for whatever experiences I might have in 
some other world, none of them (so Ayer could easily argue) would show that there 
was or was not a God as sophisticated monotheism describes Him. Of course, if God is 
conceived of as an enormously powerful being in human form, as depicted in religious 
paintings, the matter would be different, but a religious sophisticate will say that that 
is a mere image of a truth which cannot be expressed in sensory terms. At that point 
Ayer said that he was not an atheist, since "God does not exist" is as meaningless as 
"God exists." Later he relaxed this somewhat and was prepared to call himself an atheist 
on the grounds that no meaning can be given to the proposition that God exists that 
makes it remotely likely to be true. (See Ayer 1973.) 

The future of philosophy 

Officially in Language, Truth and Logic Ayer regarded the positive task left for philosophy 
- after the elimination of metaphysics and the final analysis of the statements of every- 
day life, which a verificationist approach like his had, at least almost, finally achieved 
(see pp. 152-3) - as the analysis of the concepts of science. But this was hardly under- 
taken in that work, and was never a main concern of Ayer's (except to some extent in 
his treatment of probability). His interest was always in those traditional questions of 
philosophy his radical answers to which we have been discussing, and most of his later 
work consists in attempts to find more persuasive answers to them in the light of a less 
extreme form of verificationism. 

Later positions 

It is unfair to take Language, Truth and Logic as being Ayer's main contribution to 
philosophy. There is much insufficiently admired later work. At present Ayer is out of 
fashion and undervalued. He would hardly have ground for complaint at this, scoffer 
as he was as a young man at his predecessors, but greater justice will be done to him 
one day. 



He is probably right that his best book was The Problem of Knowledge (1956). 
This presents philosophical epistemology (theory of knowledge) as primarily concerned 
to explain what knowledge is by an examination of skeptical difficulties found in 
various ordinary claims to possess it. In the process he presents a revised view of 
such matters as we have considered in connection with Language, Truth and Logic, 
revisions to a great extent already presented in the essays collected in Philosophical 
Essays (1954). 

Ayer contends that for someone to know that something is the case, it is required (1) 
that he feels sure that it is so, (2) that it is so, and (3) that he has the right to be sure 
that it is so. (See 1956: 34). 

Skeptics have raised all sorts of doubts about our right ever to be sure of the truth 
of any propositions concerning physical objects, other minds, and past events, whether 
these propositions simply assert that there really are such things or say something more 
specific about particular cases. 

The pattern of skepticism is as follows. A contrast is drawn between the evidence on 
which such propositions are believed and what they claim to be the case. Thus our evi- 
dence for propositions concerning physical objects always consists in facts about our 
own sense-data, for propositions concerning other minds in facts about the behavior of 
other organisms, and for propositions concerning past events in apparent memories or 
records of them. 

There are four types of philosophical riposte to such skepticism. 

(1) Naive realism. This denies that our evidence for the problematic knowledge 
claims is indirect in the manner the skeptic alleges, rather do we have a direct or 
immediate experience of physical things or other minds or the past (whichever is in 

(2) Reductionism. This analyses the problematic truths into truths about the 
things which feature in the evidence for them. Thus facts about physical objects 
are reduced to facts about sense-data and facts about other minds to the behavior 
of others. The interpretations of such propositions in Language, Truth and Logic are 
paradigm cases of reductionism. 

(3) The scientific approach (or causal inference theory). The problematic propositions 
are inferred as the causes of the things which are our evidence for them. 

None of these approaches is altogether successful. Statements about physical objects 
cannot be translated into statements about sense-data nor statements about other 
people's experiences into facts about their behavior. Naive realism simply ducks the 
problem. And causal realism is vulnerable to the objection that a proper causal infer- 
ence must be to a reality which we could know about more directly, and thus be able 
to check that the causal relation between our evidence and what we take it as evidence 
for really holds. 

(4) Descriptive analysis. The sting of the problem as the descriptive analyst 
sees it lies in the fact that we complain that we have not got a way of knowing 
about these things when it is logically impossible that we should have. For example, 
the complaint that other minds are closed to us loses its alarming quality when 
we realize that it is a necessary truth (analytic) that our belief in anything must 
rest on our own experience, so that it is not that we lack some power which it would 



even make sense of someone having. Once this is realized, we can be content 
with a careful description of our normal way of forming beliefs about the minds 
of others, since this is often as good as it makes sense to wish for. Similarly for the 
other problems. 

Ayer's own favored strategy in each case is the last, that of descriptive analysis. 
However, it must be said that in the case of the first problem (that of our knowledge 
of physical objects, always a special interest of his) it is doubtful how far Ayer 
really means to distance himself from reductionism. For various somewhat techni- 
cal reasons there can be no actual translation of statements about physical objects into 
if-then statements about sense-data, but the suggestion lingers in the air that 
somehow there is nothing else for them to tell us about. (See Ayer 1956: 131-47 and 
1954: essay 8.) 

Ayer's approach to the problem of other minds is somewhat different. (See 1956: 
243-54.) The obvious account of our knowledge of other minds is that it rests on 
an argument from analogy (which belongs to the causal inference type of approach 
listed above). The trouble is that the obviously respectable cases of argument 
from analogy are where the conclusion concerns the existence of something of 
the same general kind as things we have encountered more directly (say that 
there may be life on some distant planet sufficiently analogous to our own). Ayer is 
still troubled, in effect, by a verificationist scruple about anything in principle 

At one stage Ayer suggested an intriguing solution. (See Ayer 1954: essay 8, also 
1956: 247-9.) When I say of another that he is having a certain experience, what I 
mean can be analyzed into a statement of the form 'A person of such and such a 
description (e.g. presently standing in a certain position, female, capable in philosophi- 
cal argument, etc., etc.) is having such and such an experience." Now is it, Ayer asks 
us, a necessary truth that it is not I myself who answer to that description? If not, and 
it is conceivable, though profoundly contrary to fact, that I might have done so, then it 
is also conceivable that I could have verified the proposition directly. (So another's 
experience becomes more like the other side of the moon - as it seemed to be then - 
than, say, God.) 

There is something that will seem to most people rather suspicious about this solu- 
tion of the difficulty, but I cannot pursue the matter further. What it does show is that 
Ayer was never quite sure whether he continued to want statements which are mean- 
ingful for me to be ones which in principle I myself could verify, or whether it is only 
required that some human being, or like creature, could verify them. 

On all these topics Ayer developed his position further. The Central Questions of 
Philosophy, in particular, includes some quite novel suggestions on our construction of 
the physical world, which, unfortunately, we cannot consider here. 

Nor is there space here to consider the many other philosophical problems Ayer dealt 
with in his work. His discussion of probability, for example, does well to insist on some 
easily overlooked facts about it, e.g. that if statements of probability can only assert how 
probable something is relative to certain specifiable evidence then there is no way in 
which we can assert that it is more probable that a proposition based on more com- 
prehensive evidence will be true than will one based on less comprehensive evidence. 



(See 1963: 188-98 and 1972: 54-8.) His later more positive view of metaphysics, as 
conceptual clarification which may improve our way of understanding things, should 
also be mentioned. (See 1967: essay 5.) 

Altogether his philosophy, right or wrong, has admirable qualities to which 
this account, concentrating as it does, for historical reasons, on Language, Truth 
and Logic, may not have done justice. For one thing, he wrote in philosophical 
prose of unrivalled excellence. It is thoroughly straightforward and extremely lucid. 
A passage once understood stays easily in the mind as the basis for what follows 
as one reads on; there was no need for those irritating numbered propositions at 
which one must be for ever looking back in much philosophy written today. 

Ayer is for now the last great figure in the great tradition of British empiricist 
philosophy in the line of Hume. Its faults, so far as it is faulty (a matter not considered 
here), are those of that whole tradition which he may, indeed, have carried forward as 
well as it ever will be again. 


1 I shall mostly follow the usage recommended in the introduction to the second edition 
of Language, Truth and Logic, according to which any set of synonymous declarative 
(he says "indicative") sentences is spoken of as expressing a statement, and this statement 
is said to be a proposition if and only if it is literally meaningful (1946: 8). 

2 It is not altogether clear what an experiential statement or proposition is. (Ayer gives no 
examples here.) Does it report a present experience or observation or does it predict 
one expected to take place shortly? Ayer's formulation suggests the first, in which case it 
is rather pointless. Yet that seems to be what Ayer has in mind. In any case, experiential 
propositions seem very like the ostensive propositions which Ayer had rejected in the 
original first edition text (1946: 91-3). 

3 So much, indeed, is said by Ayer himself, though somewhat as an aside. "For it will be 
shown that all propositions which have factual content are empirical hypotheses; and 
that the function of an empirical hypothesis is to provide a rule for the anticipation of 
experience" (1946: 41, see also p. 151 and passim). 

It is worth noting that whereas the somewhat similarly minded American pragma- 
tists emphasized the importance of genuine factual knowledge as facilitating our control 
of things, Ayer almost exclusively speaks of prediction and I remember him once arguing 
(in a seminar) against C. I. Lewis's claim that only an agent, with some control over events, 
could understand factual statements. (For a rare use of "control" see 1946: 50.) 

4 It seems that we must either recognize a non-conventional necessity here, or agree with 
the idea many find in Wittgenstein that somehow each deduction from a rule is itself 
ultimately a free decision, or at least one necessary only in the sense of being socially 

5 Compare Stevenson 1945: 68-70. The question was sometimes raised whether the verifica- 
tion principle was synthetic or analytic. Ayer's answer was that it was analytic with reference 
to the meaning which gives what we call factual statements their point. See introduction, 
1946: 15-16. 

6 It does not follow that physical things cannot exist unperceived, since their existence consists 
not in their being perceived but in the fact that if certain conditions were fulfilled they would 
be. Thus Ayer thinks to distinguish himself from idealism. See 1946: 145. 




Works by Ayer 

1946: Language, Truth and. Logic, 2nd edn., London: Victor Gollancz. (First published 1936.) 
1954: Philosophical Essays, London: Macmillan. 
1956: The Problem of Knowledge, London: Macmillan. 
1963: The Concept of a Person, London: Macmillan. 
1967: Metaphysics and Common Sense, London: Macmillan. 
1972: Probability and Evidence, London: Macmillan. 

1973: The Central Questions of Philosophy , London: Weidenfeld andNicolson. (Reprinted Penguin 
Books, 19 76.) 

Works by other authors 

Church, A. (1949) Review of Language, Truth and Logic, Journal of Symbolic Logic 14, pp. 52-7. 
Hempel, C. G. (1953) "The Empiricist Criterion of Meaning," in Logical Positivism, ed. A. J. 

Ayer, Glencoe, IL: Free Press, pp. 115-16. 
Schlick, M. (1949) "Meaning and Verification," in Readings in Philosophical Analysis, ed. H. 

Feigl and W. Sellars, New York: Appleton-Century-Crofts, p. 158. (First published in 

Philosophical Review 45 (1936).) 
Stevenson, C. L. (1945) Ethics and Language, New Haven, CT: Yale University Press. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

J. L.Austin (1911-1960) 


John Langshaw Austin received his university education in classics at Balliol College 
Oxford. After completing his degree in 1933 he became a fellow of All Souls College 
and in 1935 a fellow of Magdalen College. During the Second World War, from 1939 
to 1945, he served as an officer in British intelligence, rising to the rank of Lt. Colonel. 
He is said to be largely responsible for the extraordinary accuracy of the Allied intelli- 
gence at the time of the Normandy invasion, and he received citations from the British, 
French, and American governments for his war work. After the war he returned to 
Oxford and in 1952 he became White's Professor of Moral Philosophy, a post he held 
until his death in 1960. 

When Austin was professor, there were about sixty practicing professional philoso- 
phers in Oxford, and only three held the rank of professor (the other two were Gilbert 
Ryle and H. H. Price). Austin was the most influential of a very distinguished group 
of Oxford philosophers of that period. During the fifties most people in Oxford thought 
it was the best university in the world for the study and practice of philosophy, and 
there was no question that philosophy was the dominant subject in the university at 
large. It is hard for people educated in other universities, even in Britain, to imagine 
the status, prestige, and intellectual centrality accorded to philosophy in Oxford at 
that time. 

The period of Austin's ascendancy matched closely my own stay in Oxford, from 
my entry as a freshman in 19 52 until, as a lecturer at Christ Church, I left in 1959. 
I got to know him quite well, and these remarks are based in part on my own personal 
recollections. Austin's influence was not primarily due to his writing. He published 
only seven articles in his lifetime, and of these only one, "Other Minds," can be said 
to have been tremendously influential at the time, though three others, "Truth," "A 
Plea for Excuses," and "Ifs and Cans" received a good deal of attention. There is a 
sense in which most of Austin's published works during his lifetime were populariza- 
tions. These were articles and lectures to meet some particular request or demand. 
Four of his articles were prepared as invited contributions to symposia of the 
Aristotelian Society and one was his public lecture to the British Academy. A sixth 
was his presidential address to the Aristotelian Society. Only one, "How to Talk - Some 
Simple Ways" was, so to speak, unprovoked. It was published as a separate article by 
the Aristotelian Society. 



During his lifetime Austin's influence was due primarily to two factors: first, he had 
an original conception of how philosophy might be practiced; and second, he had a 
forceful intellect and personality that he exhibited in his teaching, and above all in 
philosophical discussions, both with students and colleagues. His lectures on speech 
acts were published after his death, and this work is his greatest legacy, though it was 
largely unknown in his lifetime, except to people who had been his students. Now that 
he has been dead for several decades, we can appraise his contributions from a longer 
perspective, ft seems to me there are four different subjects that need to be discussed. 
First, his theory of speech acts. Second, his conception of ordinary language, and of 
ordinary language philosophy, and how it might be used constructively to give us 
greater philosophical insight. On this topic the classic work is his article "A Plea For 
Excuses." Third, Austin's conception of ordinary language philosophy and how it might 
be used critically in the examination of traditional philosophical issues. Austin's criti- 
cism of sense-data theories of perception, in his posthumously published book Sense 
and Sensibilia, is the purest expression of his critical technique. Fourth, much of Austin's 
influence both on his contemporaries, when he was alive, and on the subsequent work 
of his students and colleagues, was due to his qualities of character and intellect, f 
conclude the chapter by giving a brief assessment of his principal achievements. 

The theory of speech acts 

f believe Austin's most important contribution to the history of philosophy is in his 
overall philosophy of language as manifested in his theory of speech acts. 

During his lifetime Austin's most important discovery was supposed to be that of 
"performative utterances," and correspondingly of performative verbs and performa- 
tive sentences, fn the period in which Austin worked, philosophers generally supposed 
that the main function of language was to make truth claims. There were various ways 
of describing these, and it was common to say, as the logical positivists did, that all of 
our cognitively meaningful utterances divided into the analytic and the synthetic, and 
it was common in ethical philosophy to insist that there was a distinction between 
those utterances which were "descriptive," and those which were "evaluative." Austin 
thought that all of these simple distinctions were much too crude. He was the first 
philosopher to notice that there is an important class of utterances made with indica- 
tive sentences that do not set out to be true or false, because in these utterances the 
speaker is not describing a situation, but rather performing an action, and performing 
an action where the utterance of the sentence constitutes the performance of the action 
named by the main verb of the sentence. So if I say "I promise to come and see you," 
in appropriate circumstances, 1 am not describing a promise, 1 am making a promise. 
According to Austin, the utterance of the sentence serves to perform an action, not to 
describe anything. This led him to make a distinction that he thought would enable us 
to see matters more correctly: the distinction between performative and constative 
utterances. There are three ways in which performatives differ from constatives. First, 
performatives such as "I promise to come and see you" typically have a special verb for 
performing the action in question, and there is even a special adverb, "hereby," which 
we can insert in performative sentences; for example, "I hereby promise to come and 
see you". Constatives, for example, "ft is raining," or "Snow is white," do not have or 



need a special verb. Second, constatives can be true or false, but performatives are 
not true or false, rather they are either felicitously or infelicitously performed. 
Corresponding to the true/false dimension for assessing constatives is the felicitous/ 
infelicitous dimension for assessing performatives. And third, the performative utter- 
ance is an action, a doing, whereas the constative is a statement or a description. 

However, Austin's patient research eventually showed that this way of making the 
distinction does not work. It turns out that all of the features that are supposed to be 
special to the performative are true of the constative as well, and thus what was origi- 
nally supposed to be the special case, performatives, seems to swallow the general case, 
constatives, which now turn out to be performances of actions like any other utterance, 
and this led Austin to a general theory of speech acts. 

Going through the three criteria in order: first, just as there are performative verbs 
for promising, ordering, and apologizing, so also there are performative verbs for 
stating, claiming, and other constatives. Thus, just as one can promise by saying "I 
promise," so one can state that it is raining by saying "I state that it is raining," and the 
criterion that Austin had hoped to use to identify performative verbs, namely the 
possible occurrence of the adverb "hereby," as in "I hereby promise to come," also 
characterizes constatives, as in "I hereby state that it is raining." 

Second, the so-called constatives also have a felicitous/infelicitous dimension of 
assessment, and many so-called performatives can be appraised as true or false. For 
example, if I make a statement that I am no position to make, my utterance will be infe- 
licitous in exactly the same sense that a promise can be infelicitous if, for example, I am 
unable to do the thing I promised to do. Suppose I say right now, "There are exactly 
thirty-five people in the next room," when I have no basis whatever for making that 
statement, then the statement is infelicitous in the same sense in which performatives 
can be infelicitous. Furthermore, there clearly are apparent performatives that can be 
judged as true or false. If I say, "I warn you that the bull is about to charge," when it is 
not the case that the bull is about to charge, then I have issued a false warning, even 
though a warning is a performative on Austin's original definition. 

Third, making a statement is just as much performing an action as making a 
promise. At the end of Austin's discussion the conclusion is obvious: we should think 
of every utterance as the performance of a speech act. The notion of a performative 
should be restricted to those utterances containing the performative use of a 
performative expression. 

The theory of speech acts begins with the rejection of the performative/constative 
distinction. Within the theory of speech acts Austin then made a distinction between 
three different levels of description of an utterance: (1) the level of the locutionary act, 
which is defined as uttering words with a certain meaning, where "meaning" is 
explained as sense and reference; (2) the level of the illocutionary act, which is defined 
as the utterance of words with a certain force, which Austin baptized as "illocutionary 
force"; and (3) the perlocutionary act, which is defined as the production of certain 
sorts of effects on the hearer. To take Austin's example, if I say "Shoot her," then if by 
"shoot" I mean shoot, and by "her" I refer to her, then I will have performed a certain 
locutionary act of saying "Shoot her." But if I uttered that sentence with the force of 
an order, or advice, or request, then those verbs will name the illocutionary force of my 
utterance and hence the illocutionary act that I was performing in making the utter- 



ance. And if I persuade the hearer to shoot her, persuading is the production of an effect 
on a hearer of a sort that Austin called a perlocutionary act. 

The distinction between the illocutionary and the perlocutionary seems to me essen- 
tial for any theory of language, and it is especially important for those theories that take 
language as a matter of linguistic behavior; because, of course, the linguistic behavior 
which involves producing effects on people in the form of perlocutionary effects, needs 
to be distinguished from the linguistic behavior which involves performing speech acts, 
regardless of the subsequent effects on the hearers. Implicit in Austin's work is the con- 
ception that the illocutionary act, not the perlocutionary act, is the fundamental target 
of analysis in the philosophy of language. That has been the assumption on which I 
and a large number of other researchers have proceeded in attempting to carry on 
Austin's pioneering efforts (see SEARLE). 

The distinction between the locutionary and the illocutionary, however, does not 
seem to me to work. The reason is that the meaning of the sentence, which is supposed 
to determine the locutionary act, is already sufficient to fix a certain range of illocu- 
tionary forces. You cannot distinguish between meaning and force, because force is 
already part of the meaning of the sentence. There is no way that I can utter the 
sentence "It is raining," or for that matter, "Shoot her," without performing some 
illocutionary act insofar as it is a locutionary act. There is no distinction between the 
locutionary and the illocutionary, because the locutionary is eo ipso illocutionary. 

Austin also gave us a taxonomy of types of illocutionary acts. His taxonomy includes 
the following; first, verdictives. These are findings of fact or value on some matter. An 
example of a verdictive is giving a verdict. Second, exercitives. These are the exercising 
of powers, rights, and influence. Examples would be appointing and voting. Third, com- 
missives. These are always cases of committing the speaker to a course of action. The 
favorite example, of course, is promising. Fourth, behabitives. These have to do with 
social behavior. Examples are apologizing and congratulating. Fifth, expositives. These 
make plain how our utterances fit into the discourse. Examples are replying, arguing, 
and conceding. 

As with the locutionary/illocutionary distinction, it seemed to me this taxonomy 
needs revision and extension, because there is no clear criterion for distinguishing 
between the various categories. I and several other philosophers have attempted to 
criticize and improve on Austin's taxonomy; however it is important to emphasize that 
the criticisms and revisions of his views are made within a framework he invented 
and using tools he gave us. I see the many criticisms of Austin's specific doctrines by 
subsequent speech act theorists not as refutations but as further contributions to a 
discussion that he began, but did not live to complete. 

It is important to emphasize that when we read Austin's most famous book, the work 
posthumously published, How To Do Things With Words, we are reading his lecture notes. 
Austin would never have published this material in this form. I know this for a fact 
because I wanted him to publish it so that I could publish my criticisms of it, even when 
I was a student. I once asked him "How soon can we hope that your William James 
lectures will be published?" thus giving him an opening I should never have done. He 
responded immediately, "You can hope it will be published any time you like." Further 
discussion revealed that he did not think the work ready for publication: "It is too half- 
baked," he said. 



Ordinary language philosophy: the constructive function 

Austin was most famous during his lifetime, not for his theory of speech acts, about 
which only the theory of performatives was generally known, but rather for his par- 
ticular conception of philosophy, and his style of doing philosophy. He was always 
anxious to insist that he did not think that his was the only way of doing philosophy, 
but merely that it was one possible way of doing one part of philosophy. He thought 
that the first step to be taken in philosophy was to make a very careful analysis of the 
ordinary use of expressions. The ordinary expressions of a natural language like 
English, he thought, embodied all the distinctions about the world that people had 
found it necessary and useful to make in the course of millennia. He did not think that 
ordinary language was the last word, but he did think it was the first word. In a debate 
with Bertrand Russell, when Russell asked him if he thought the examination of ordi- 
nary language was the be-all and the end-all of philosophy, Austin is reported to have 
answered, "It may not be the be-all and the end-all, but it certainly is the begin-all." 
The analysis of the ordinary use of expressions served two philosophical purposes 
for Austin. One was a corrective purpose of showing that many of the claims that 
philosophers had made rested simply on mistakes about the ordinary use of expres- 
sions. His most famous discussion in this regard is probably his criticism of the 
arguments for the sense-data theory of perception, in his lectures Sense and Sensibilia. 
The second purpose of the analysis of language was more constructive: he thought we 
could learn a great deal about the world by analyzing the expressions we use to describe 
the world. 

Austin thought that his method of doing philosophy allowed for two features which 
philosophy is thought not to possess. First, philosophy on his conception is a coopera- 
tive enterprise. It is not something you do alone in your study, but rather you get a group 
of people and try to discuss examples to see how words are used in describing those 
examples. And second, philosophy so construed allows for progress. It is typical that 
people who carry on philosophical discussion of this type, at the end of the day, feel 
they have made definite progress in analyzing the application of words to concrete 

It is characteristic of Austin's approach that he can often show that what seem like 
two synonyms or near synonyms are really quite different. In a famous case he took 
the two expressions "by accident" and "by mistake" and showed that they really had 
quite different meanings, even though at first sight most English speakers would prob- 
ably say they mean pretty much the same thing. Here was his demonstration. Suppose 
I go out into the field to shoot my donkey. Suppose I see your donkey, which looks very 
much like my donkey, and I shoot your donkey. Did I shoot your donkey by accident, or 
by mistake? Suppose I go out into the field and shoot at my donkey, but just as I am 
pulling the trigger, the two beasts move, and my bullet strikes your donkey. Did I shoot 
your donkey by accident, or by mistake? I think the examples are absolutely clear in 
both cases. 

The constructive side of Austin's method of doing philosophy is most powerfully 
exemplified by his article 'A Plea For Excuses" (together with its posthumously pub- 
lished companion piece, "Three Ways of Spilling Ink"). 'A Plea for Excuses" is, in fact, 
a summary of an entire series of seminars that Austin gave during the 1950s. I 



attended the seminars, and there was easily enough material presented to fill an 
entire book, but it is perhaps typical of Austin that the material that a more average 
philosopher would use for a complete book he condensed into a single article. Austin's 
method is illustrated by the following: most philosophers, myself for example, if exam- 
ining the problem of action, would begin by asking what fact about an event makes it 
into a human action. Austin thinks that this approach, as he frequently said in 
criticizing my views, is "much too fast." His own approach is, so to speak, to sneak up 
on the problem indirectly by asking what sorts of excuses, justifications, extenuations, 
and explanations we offer for our actions. "Excuses," he insists, is just a title, not a 
description of the whole subject matter. The results of the analysis are a series of theses 
that he advances about the character of our conceptual apparatus for discussing 
actions. Many of these are quite surprising. So, for example, I think most philosophers 
intuitively would suppose that any action is done either voluntarily or involuntarily. But 
Austin points out that the whole question of negations and opposites is much more 
complex than that. The opposites of the word "voluntarily," he says, might be "under 
constraint" or "duress" or "obligation." The opposite of "involuntarily" might be 
"deliberately" or "on purpose" or the like. Austin urges us not to take anything for 
granted about negations and opposites. Again, f think many philosophers suppose that 
there is not much difference between doing something intentionally, deliberately, and 
on purpose. Austin makes it abundantly clear that these are not at all the same. He also 
urges us to pay close attention to legal cases and psychological studies; he examines 
one case, Regina v. Finney, in some detail, showing that the lawyers and the judge make 
serious mistakes, treating several terms of excuse as equivalent when they are not, 
and being unclear about what action exactly of the defendant is being qualified by 
what expression. It is impossible to summarize this article because the article is itself 
a summary of a quite extended project of research, and the interest of the results 
is in the specific details. But the article reveals both the strengths and some of the 
limitations of Austin's method. 

Ordinary language philosophy: the critical function 

I believe the purest case where one can observe Austin, so to speak "in action," is in his 
book Sense and Sensibilia. The actual text that we have before us now was prepared from 
notes of numerous students by Geoffrey Warnock, but Warnock does an excellent job 
of conveying the flavor of the actual lectures, as I can say from having attended them. 
If Austin had lived, I doubt that he would ever have published these lectures as they 
stand. Their results are almost uniformly negative, and the tone is often more harsh 
than Austin would normally have allowed in publication. Nonetheless, they are a beau- 
tiful exemplification of his method of philosophical analysis. He simply does a careful 
word-by-word examination of a series of traditional philosophical arguments designed 
to show that we never perceive "material objects," but only perceive "sense-data." 
Austin takes Ayer's book The Foundations of Empirical Knowledge as his "stalking horse," 
and he also discusses arguments from Price and Warnock (see AYER). Austin goes 
patiently through the arguments that are traditionally called "the argument from 
illusion," which attempt to prove that all we ever perceive are sense-data, and he shows 
that without exception the arguments, as presented by Ayer, are hopelessly muddled 



and confused. Ayer assumes that such words as "look," "appear," and "seems" can be 
used indifferently as if they meant the same thing, but Austin's patient analysis shows 
that they are really quite different. In the standard arguments for sense-data Austin 
finds only carelessness, muddle, and confusion. 
The stages of the arguments that he finds are 

1 The philosophers assume that there are two exclusive classes of sense experience, 
those of material objects, and those of sense-data. 

2 They argue that there must be no discriminable differences in the character of the 
perceptions since we can confuse one thing for the other. 

3 They conclude that, since one would expect a considerable difference from two such 
different sorts of entities, there must be only one class that we are actually 
perceiving, and that one must be sense-data. 

By patiently working through the texts Austin challenges each of these claims. 

1 There are all sorts of things we perceive that do not fit either category comfortably, 
such things as shadows, clouds, gases, flames, rainbows, images, etc. Austin thinks 
the dichotomy is a typical philosophers' oversimplification. It would be just as 
confused to say that all we perceive are material objects as it is to say that all we 
perceive are sense-data. 

2 In real life there are all sorts of differences in the character of our experiences. 
Dreams, for example, are different from waking experiences in all sorts of ways; and 
even the stock-in-trade waking-life examples of the epistemologist are misdescribed. 
For example, the stick in water which "looks bent" does not look like a bent stick 
out of water, and even in water it need not look to be bent. 

3 It does not follow from the arguments as presented that all we ever see are sense- 
data, and indeed it is quite arbitrary for the philosophers to select "sense-data" or 
"material objects" as the objects of perception. 

It is important to emphasize that in criticizing the sense-data theory, Austin is not 
defending the idea that all we see are material objects. He thinks that idea is just as 
crude as its opposite. 

In the course of his discussion he introduces the idea of what he called a "trouser 
word," and what some philosophers subsequently came to call "excluders." Some 
words get their meaning in a context from the words that they are opposed to in that 
context. Thus real cream is opposed to artificial cream, but a real duck is opposed to a 
toy duck or a decoy duck, and real teeth are opposed to false teeth. The word "real" is 
an excluder that gets its meaning in context from what it is opposed to. There is no 
common property of reality which the word "real" invariably and literally serves to 
ascribe. A decoy duck for example, though not a real duck, may nonetheless be a 
real decoy, as opposed for example to a paper model of a decoy duck. When the episte- 
mologist talks about reality and perceptions of reality, he fails to appreciate the nature 
of the concept. 

In a reply to Austin, Ayer claimed that his main points could survive, even if he 
accepted all of Austin's specific objections. His main point is that we could have all of 
the experiences we do have and still be mistaken in our claims about objects and states 
of affairs in the world. To this I think Austin would have replied, first that this does not 



show that all we ever perceive are the experiences, the "sense-data." And, second, it 
does not show that the relation between the experiences and the objects they are taken 
to be experiences of, is one of evidence. It does not show that the experiences are 
evidence for the presence of the objects. 

Whatever the merits of the debate, I think Austin's critique proved immensely influ- 
ential historically. One used to hear a lot about the sense-data theory of perception; one 
does not hear much about it anymore. 

Other works 

Austin wrote a number other important works which limitations of space prevent me 
from exploring in any detail, but I must mention them in passing. "Other Minds" pre- 
sents a criticism of traditional epistemology which is very much in the spirit of Sense 
and Sensibilia. "Truth" and "Unfair to Facts" present a version of the correspondence 
theory of truth and a response to Strawson's criticisms of it. I think Austin is right that 
the fundamental notion of truth is correspondence, but the particular version that he 
presents does not survive Strawson's objections. "Ifs and Cans" is a gem of philosoph- 
ical analysis. As far as the history of philosophy is concerned, its main point is to 
respond to those versions of compatibilism about the free will problem which maintain 
that to say that I could have done otherwise just means that I would have done other- 
wise if I had so chosen, and to say that I can do something just means that I will do it 
if certain other conditions are met. But Austin makes a large number of other points 
about related conceptual issues. 

Character and intellect 

My own impressions of Austin are somewhat different from those of many people who 
thought of themselves as his close associates and followers. One trait that we would all 
agree on was his immense carefulness and precision. Not only when doing philosophy, 
but even in the most casual conversation, Austin spoke and thought with great preci- 
sion, and he did not tolerate looseness in his students or colleagues. The worst con- 
demnation that he could make of something he was reading would be to shake his head 
sadly and say in his thin, precise way, "It's just loose." Indeed, on several occasions he 
said to me in tones more of sadness than anger, "There is a lot loose thinking in this 
town." For the most part, Austin's colleagues regarded him with a kind of awe, and it 
seemed to me that to some extent they were even terrified in his presence. Certainly, his 
presence in seminars and meetings had a profound effect on the behavior of others par- 
ticipating. I noticed this when I went back to Oxford some years after his death and 
found that many of the professional philosophers were behaving like schoolboys during 
recess. They were much less cautious than they would have been in Austin's presence. 
At first I could not understand the source of Austin's influence, because it seemed 
to me that I could I beat him in argument; and like a lot of undergraduates I thought 
the test of a philosopher was how good he was in the give and take of philosophical 
repartee. Austin's technique in discussion was always to take everything dead literally, 
and then to insist on certain linguistic distinctions that he thought were being over- 
looked. So, for example, when Wittgenstein's Philosophical Investigations was published, 



I and some other undergraduate philosophers insisted that we discuss Wittgenstein's 
private language argument in Austin's informal instruction for undergraduates. 
Austin's technique was to refuse to grant Wittgenstein any leeway at all. At one point 
when we were discussing Wittgenstein's famous example of the beetle in the box, 
Austin said sarcastically, 'All right, for our next session everyone bring a box with a 
beetle in it." At one point in Wittgenstein's discussion, he says there might be nothing 
at all in the box. Austin thought Wittgenstein was simply contradicting himself. "First 
he says, there definitely is a beetle in the box, and then he says there might be nothing 
in the box, a plain contradiction." 

Austin's habit of insisting on the highest level of precision, both in his professional 
activities and even in ordinary conversations, seems to me one of the main reasons for 
the terror that he inspired in his colleagues. At the time, much of the source of Austin's 
influence derived from his schoolmasterly style. Most Oxford philosophers of the time 
had been students at British boarding schools, and Austin was, so to speak, the ulti- 
mate schoolmaster. If he were reading a paper that I had written, a typical question he 
put to me would be, "Why exactly did you use the subjunctive?" Or, on another occa- 
sion, "In the verb 'suppose,' what does the 'sup' mean?" Austin was famous at the time 
for his attention to the minute details of ordinary language, but it seemed to me then, 
and it seems to me now, that his real contribution to philosophy was not so much in 
the details. Austin did indeed have a genius for spotting linguistic differences and dis- 
tinctions where most people would have thought there were none, though in the details 
his views were sometimes mistaken. His most important contribution to philosophy, I 
believe, is in his overall vision of language. 

Though he was regarded as terrifying by many of his colleagues, I can say that 
to undergraduates he was immensely kind, patient, helpful and, in his reserved way, 
even friendly. His contempt was reserved for people he thought of as pretentious, 
self-important, pompous, and above all obscurantist. When Austin, along with several 
other Oxford philosophers and students, went off to Royaumont in France for an 
English-French philosophy colloquium he considered the pretentiousness of Merleau- 
Ponty then the most influential French philosopher, quite ridiculous. "That Merleau- 
Ponty he is just a little tin god. He will never get anywhere." Austin would have 
hated the "deconstructionists" and "postmodernists" who currently pretend to admire 
his work. 

He did not think his brand of philosophy was the only correct way to do the subject, 
but he did try to extend its influence with an almost missionary zeal. Again, when we 
were in Royaumont, and he saw me in discussion with an elderly distinguished French 
philosopher, he took me aside and said, "Don't waste your time on the aged. Talk to the 
young!" This remark annoyed me at the time because I did not think of myself there 
for any other reason than to practice philosophy. Austin, I believe, would have thought 
it a waste of time for us to go to France if we did not try to spread the truth. 

I never heard anyone other than his wife address him by his first name, and when 
one of his colleagues had the temerity to address him as "John," Austin is reported to 
have said evenly, " Austin' is also a Christian name." He did have a habit of holding 
one's attention in discussion. So, for example, when making an involved point he would 
take his pipe in one hand and light a match with the other. Never taking his eyes off 
his interlocutor, he would allow the match to burn ever closer to his bare fingers until 



at the last millisecond he would flick his wrist to extinguish the match, whereupon he 
would start the whole process over again while continuing his relentless discourse, eyes 
always on his listeners. 

Austin's reluctance to publish was part of the culture of Oxford at the time, but also 
it was partly characteristic of his own attitudes. Oxford had a long tradition of not pub- 
lishing during one's lifetime, indeed it was regarded as slightly vulgar to publish. People 
who did publish a lot, like A. J. Ayer, were regarded as remiss for having published too 
much too soon. As far as having a career and making a reputation were concerned, the 
attitude in Oxford was that the only opinions that really matter are the opinions of 
people in Oxford, and perhaps a few in Cambridge and London, and they will know 
about one's work anyway. One does not need to publish. What one does not want is a 
lot of graduate students somewhere, picking over one's half-baked published texts and 
- horror of horrors - finding mistakes. So I think Austin's reluctance to publish was 
partly due to his extreme carefulness, but it was partly due to his sheer vanity; he did 
not want any intellectual inferior pointing out errors. 

At a time when anti-Americanism was very common in Britain, especially among 
the intellectual classes, Austin simply adored the United States, and especially its 
university system. He would not tolerate criticisms of the United States, and the only 
subject on which I have ever heard him show uncritical enthusiasm was America. 
Indeed, he once said to me, "The future lies with America," and on another occasion, 
"There are unplowed fields in that country." Once Herbert Hart was criticizing 
American cooking, and Austin said evenly, in his discussion-ending way, "It is not 
so bad." 

I often read how much Austin was influenced by Wittgenstein. Nothing could be 
further from the truth. Austin had no sympathy whatever for Wittgenstein, and I think 
he was incapable of learning from someone whose style was so "loose." He typically 
referred to Wittgenstein in the style of English schoolboy slang of the time as, "Witters," 
pronounced "Vitters." He thought there were no original ideas in Wittgenstein. Indeed 
he once said to me about Wittgenstein's philosophy, "It's all in Moore," one of the least 
accurate things I have ever heard Austin say. If Austin had an inspirational model, it 
was Moore (see MOORE). 

It will seem a paradoxical feature of Austin's career that he aroused such passion- 
ate controversy both pro and con. On the surface, at least, his presentations are invari- 
ably modest, cautious, and self-effacing. Though I did not regard myself as one of his 
followers, I found it easy to see why they regarded him with such enthusiasm. He 
offered them a new conception of philosophy, and with it, a new research program. But 
it is more puzzling to try to understand why he was hated so much. I think in order to 
understand the hostility that he aroused, we have to compare his career with that of 
Socrates. He was hated for much the same reason that Socrates was hated: he seemed 
to destroy everything without leaving anything substantive in its place. Like Socrates 
he challenged orthodoxy without presenting an alternative, and equally comforting, 
orthodoxy. All Austin offered, again like Socrates, was a new method for doing 

Austin's substantive achievement, especially in the theory of speech acts, has sur- 
vived his death now for nearly a half century, and will, I believe, continue to be a focus 
of research. But his official doctrine as to how philosophy might be pursued has waned 



considerably. It has very few followers and practitioners. Why? Well, part of the answer 
is that it is just too difficult. The sort of very careful analysis of minute linguistic 
distinctions that Austin urged us to undertake is simply too much work for most 
philosophers. Austin thought that we ought to be more patient and hardworking. If 
entomologists can classify a million different kinds of insects, surely philosophers ought 
to have the patience to classify the few dozens or few hundred or even a few thousand 
different sorts of uses of different sorts of words. But the problem is that the motiva- 
tion that tends to make one a philosopher seems to be quite different from the motiva- 
tion that makes one an entomologist. Philosophers want very general answers to very 
large questions, whereas Austin thought they had first better get clear about the 
distinctions among a number of adverbs, working themselves up to undertake an 
analysis of a few verbs. 

There are certain limitations on Austin's methods, which, paradoxically but to his 
credit, we can use Austin's theory of speech acts to expose. 

(1) Sometimes Austin confuses the truth conditions of a term, that is, the condi- 
tions under which it is a fact that some object satisfies that term, with the conditions 
for appropriately asserting that the term applies. Thus, to take an Austin-style example, 
Austin points out that we wouldn't normally assert that a man walked across the room 
intentionally, if he just walked across the room in an ordinary, unexceptional way. "No 
modification without aberration," Austin tells us. Nonetheless it may be true that 
the man walked across the room intentionally, it is just not appropriate to assert it 
unless there is something unusual. It may just be too obvious that the act was done 

(2) Related to the first mistake is the mistake of confusing the meaning of a term 
with the illocutionary force that characteristically accompanies the assertion that 
the term applies to an object. Thus, to take another example from Austin, he points out 
that when we say that we know something, we are often giving our guarantee for what 
we claim to know, that the claim to know has certain features in common with a per- 
formative, such as "I promise." Austin is careful not to say that "know" is a performa- 
tive, but he does think that the assertion that one knows has a performative-like 
guaranteeing force. But once again this does not tell us the meaning of the word 
"know," because it cannot account for the occurrence of this word in other cases 
such as conditionals or negations. So even if an utterance of the form "I know that p" 
means something like "I guarantee thatp," still, an utterance of a conditional of the 
form "if I know that p then q," does not mean anything at all like "if I guarantee that 
p, then q." 

(3) Even after you have done a careful linguistic analysis and shown that the 
standard philosophical positions rest on a misuse of words, still, you can often state the 
position again without using those words. The problem remains even after the misuse 
of words have been corrected. Thus, to take the problem of free will, Austin points out 
that when we say that an act was done freely "freely" functions as an excluder, exclud- 
ing all of the various ways in which an act may not have been done freely, such as, for 
example, done under duress or under compulsion. But even if Austin is right about this, 
and he probably is, you still have a free will problem left over. Here it is: Are all human 
acts such that the performance of the act has antecedent causal conditions which are 
causally sufficient to determine the act? If I walk across the room, and I walk across 



the room in a way which is not under duress or compulsion, all the same there is still 
the question about free will. Were the antecedent conditions prior to the onset of my 
action of walking across the room sufficient to determine that 1 was going to walk 
across the room? That question remains even after we have become clear about all of 
the various uses of freely, voluntarily, etc. 

(4) Austin says that ordinary language embodies all of the distinctions that 
humans have chosen to make over millennia. But there is a sense in which that is not 
quite right. We can indeed state in ordinary language all of the distinctions that 
humans have chosen to make, and indeed we can state a lot that they have not chosen 
to make. But it is not the case that every real distinction that humans have made is 
marked by a lexical distinction, by two different words, in ordinary usage. Thus to take 
one of Austin's examples, the word "pretend" does not mark a distinction between those 
cases of pretense which are genuinely intended to deceive, and those cases of pretense 
which are put on or are mock-performances, but not designed to deceive. So if I pretend 
to be the President of the United States in order to be admitted to the White House, I 
have pretended in the deceptive way. But if f pretend to be the President of the United 
States as part of a game of charades, there is no intention to deceive. This is an obvious 
and important distinction, as Anscombe pointed out in the symposium with Austin on 
pretending, but we do not have two verbs whose meanings are "pretend deceptively" 
and "pretend non-deceptive ly." 


J. L. Austin was one of the most important philosophers of the twentieth century. In 
examining his contribution we need to distinguish between the philosophy of language 
and linguistic philosophy. The philosophy of language is the attempt to give an account 
of certain very general features of the structure, use, and functioning of language. 
Linguistic philosophy is the attempt to solve philosophical problems by using linguistic 
methods. Austin made important contributions to both the philosophy of language and 
linguistic philosophy. During his lifetime he was famous as a linguistic philosopher, but 
not for his philosophy of language. Since his death it has emerged that his most impor- 
tant contribution to philosophy has been his philosophy of language as expounded in 
his theory of speech acts. At the conclusion of "Ifs and Cans," Austin expresses the hope 
that the next century may see the birth of a comprehensive science of language. I believe 
that he thought his theory of speech acts was a contribution toward that future science. 


Works by Austin 

1962: How To Do Things With Words, Cambridge, MA: Harvard University Press. 
1964: Sense and Sensibilia, Oxford: Oxford University Press. 

19 79: Philosophical Papers, 3rd edn., Oxford: Oxford University Press. (2nd edn., 19 70.) Articles 
in this collection are given below, with first publication details where applicable: 
"Agathos and Eudaimonia in the Ethics of Aristotle," from Aristotle: A Collection of Critical 
Essays, ed. J. M. E. Moravcsik, New York: Doubleday, 1967, and London: Macmillan, 1968. 
"A Plea for Excuses," Proceedings of the Aristotelian Society 5 7 (1956-7). 



'Are There A Priori Concepts?," Proceedings of the Aristotelian Society, Annual Conference, 

"How to Talk," Proceedings of the Aristotelian Society 53 (1952-3). 
"Ifs and Cans," Proceedings of the British Academy 42 (1956). 
"Other Minds," Proceedings of the Aristotelian Society, suppl. vol. 20 (1946). 
"Performative Utterances." 

"Pretending," Proceedings of the Aristotelian Society, suppl. vol. 32 (1957-8). 
"The Line and the Cave in Plato's Republic." (Not in 1st and 2nd edns.) 
"The Meaning of a Word." 

"Three Ways of Spilling Ink," Philosophical Review 75/4 (1966). 
"Truth," Proceedings of the Aristotelian Society, suppl. vol. 24 (1950). 
"Unfair to Facts." 

Works by other authors 

Anscombe, E. (1957-8) "Pretending," Proceedings of the Aristotelian Society, suppl. vol. 32. 

Ayer, A. J. (1940) Foundations of Empirical Knowledge, London: Macmillan. 

(1973) "Has Austin Refuted Sense-data?," in Essays on ]ohn L. Austin, ed. I. Berlin et al., 

Oxford: Clarendon Press. 
Fann, K. (ed.) (1969) Symposium on J. L. Austin, London: Routledge and Kegan Paul. 
Holdcraft, D. (19 78) Words and Deeds, Oxford: Oxford University Press. 
Searle, J. (1969) Speech Acts: An Essay in the Philosophy of Language, Cambridge: Cambridge 

University Press. 
(19 79) Expression and Meaning: Essays in the Theory of Speech Acts, Cambridge: Cambridge 

University Press. 
(1989) "How Performatives Work," Linguistics and Philosophy 12, pp. 535-58. 

Travis, C. (19 75) Saying and Understanding, Oxford: Blackwell Publishers. 

Warnock, G. (1991) /. L. Austin, Arguments of the Philosophers series, New York: Routledge. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

Norman Malcolm (1911-1990) 



Norman Malcolm was born on June 11, 1911, in Selden, Kansas, and died in London 
on August 4, 1990. His undergraduate years were at the University of Nebraska, where 
0. K. Bouwsma was one of his teachers. His Ph.D., granted in 1940, was from Harvard, 
but the most important philosophical influences on him during his graduate years were 
G. E. Moore and Ludwig Wittgenstein, with whom he studied during a fellowship at 
Cambridge University in 1938-9. He was an instructor at Princeton before joining the 
US Navy in 1941. After the war he spent another year in Cambridge, 1946-7, study- 
ing with Moore and Wittgenstein. In 1947 he joined the Sage School of Philosophy at 
Cornell, where he remained until his retirement in 1978. During the last twelve years 
of his life he lived in London and was appointed a Visiting Professor and Fellow at King's 
College London, where he gave a weekly seminar mainly devoted to the philosophy of 

Malcolm credited Moore with being the first to employ the technique of refuting 
paradoxical philosophical statements by pointing out that they go against ordinary 
language - that they imply that ordinary uses of language are incorrect uses - which 
is Malcolm's own favorite technique. From Wittgenstein he took the idea that a philo- 
sophical problem is essentially a confusion in our thinking that is to be remedied by 
reminders of the actual use of language, and by reconstructing and criticizing the 
analogies and reasoning that bewitch the victim of the puzzle. 

Malcolm was a major expounder and endorser of Wittgenstein's later philosophy. He 
devoted several articles explicitly to explaining Wittgenstein's thought. The earliest and 
probably most influential of these was his discussion of Wittgenstein's Philosophical 
Investigations published in The Philosophical Review in 1954, which prompted a good 
deal of interest in Wittgenstein's argument against the possibility of a private language. 
Wittgenstein visited Malcolm in Ithaca in 1949 and their discussions there of know- 
ledge and certainty stimulated the thinking that led Wittgenstein to his last major work, 
On Certainty. Late in his life, in 1986, Malcolm published a book, Nothing is Hidden, the 
aim of which is to expound Wittgenstein's later criticism of his Tractatus Logico- 
Philosophicus. And at his death he left a monograph, Prom a Religious Point of View? 
(published posthumously, edited and with a response by Peter Winch) in which he 



summarizes much of Wittgenstein's work, early and late, in an attempt to see in what 
sense Wittgenstein's remark that he approached every problem from a religious point 
of view might be true. But nearly all of Malcolm's work, from the early 1950s on, is 
shot through with approving reference to remarks of Wittgenstein's. It would not be 
an exaggeration to say that he aimed nearly all of his work at getting across insights 
he owed to Wittgenstein (see WITTGENSTEIN). 

Malcolm's writing is remarkable for its clarity and vigor and its freedom from tech- 
nical jargon. It is crammed with down-to-earth examples. These sometimes help to give 
a concrete grasp of an abstract idea, but typically they serve to remind his readers of 
how ordinary language is actually used. They always give his writing considerable 

What follows are brief expositions of some of the views Malcolm argued for (I find 
no evidence in his writings of any major change in his views), placed under four head- 
ings: knowledge, mind, memory, and philosophy of religion. This sample is far from 
comprehensive, but I hope it gives a good idea of the breadth and character of 
Malcolm's work. 


In two important early papers, "Certainty and Empirical Statements" (1942) and "The 
Verification Argument" (1950), Malcolm rebutted the claim made by some philoso- 
phers (e.g. C. I. Lewis, Carnap, Russell, Ayer) that it is impossible for an empirical state- 
ment (a contingent statement about material objects) to be known with certainty. In 
Lewis and Carnap he finds an argument for this paradoxical claim, which he calls the 
Verification Argument. He arrives at the following formulation of the argument ("The 
Verification Argument", in 1963a: 26): 

I. Any empirical statement S has consequences (not in the sense of entailment but 
in the sense in which it is a consequence of "Yesterday the phrase 'the stream 
of thought' was on page 224 of vol. I of my copy of James's The Principles of 
Psychology" that if I were to look on that page now I would see that phrase). 

II. The consequences of S are infinite in number. 

Ilia. It is not certain that the consequences of S will occur. 

IVb. If any empirical statement can be conclusively established as true or false, then 
if a sufficient number of the consequences of S should fail to occur then it would 
be absolutely conclusive that S is false. 

Va. If at any time it should be absolutely conclusive that S is false then at no previ- 
ous time did anyone make absolutely certain that S is true. 

It does follow from these premises that, for any empirical statement S, no one ever made 
absolutely certain that S is true. Malcolm denies premise Ilia. He argues that it was 
accepted by proponents of the Verification Argument only because they thought it 
follows from III: It is possible that the consequences of S will fail to occur. But they 
thought this only because they failed to distinguish among different interpretations of 

III. For those senses of "possible" in which III is true ("The consequences of S will fail 
to occur" is not self-contradictory; no consequence of S is entailed by the grounds for 
holding it true), III does not entail Ilia; and for those senses of "possible" in which 



III entails Ilia (there is some reason to believe that the consequences of S will not 
occur; there is no reason to think that the consequences of S will occur; the grounds 
for holding that the consequences of S will occur are not absolutely conclusive), III is 
not true. 

In his earlier work on knowledge, Malcolm seemed to share the common assump- 
tion that statements of the form "S knows thatp" state a fact about S, a fact about whose 
necessary components philosophers might hope to say something informative, e.g., that 
S must believe p, that p must be true. But in many of his later writings, Malcolm seems 
to treat such sentences, particularly in the first-person present-tense form, as like per- 
formatives whose use is, not primarily to report a fact about the subject, one that is 
there independently of any utterance, but rather to achieve some aim of the speaker, 
so that the correctness and intelligibility of its use depend heavily on the context of the 
use. (There were, however, already intimations of this idea in early papers. In 
"Defending Common Sense" (1949) he said that Moore misused "know" in making 
such assertions as "I know that I am a human being" or, when holding up his hand, "I 
know that this is a hand," in the contexts in which he made them because there was 
not any doubt or disagreement about the matter that would give a point to such asser- 
tions. In "Philosophy for Philosophers" (1951), for such reasons as that a sentence like 
"I know I feel hot" is almost never seriously used, that the normal usage of "I know" is 
informative and connected up with investigating, finding out, making sure, producing 
evidence, with asking and answering "How do you know?", he said that "In the sense 
of 'knowledge' in which knowledge is contrasted with belief, we do not (and cannot) 
have knowledge of our own sensations" (p. 336).) 

His 19 76 paper "Moore and Wittgenstein on the Sense of 'I Know'," says that "I 
know" does a variety of jobs in ordinary language use; for example, "it is used to claim 
the possession of evidence, or expertise, or ability; it is used to comfort, reassure, express 
agreement; it is used to say that one has thoroughly checked something, or that one 
can be relied on, or that one doesn't need to be reminded" (1977b: 192). For such 
reasons he often says that, except in very special contexts, it makes no sense to say "S 
knows that he is in pain" and what it means when it does make sense is very different 
from what is meant by "S knows that there is a gash in his hand." He refuses to accept 
that there is in this case any distinction (such as has been suggested by Grice, Searle, 
and others) between truth-conditions and requirements for the aptness of asserting (see 
GRICE and SEARLE). It is unclear what his view would be about whether there is such a 
distinction for such sentences as "Moore is a human being" or "What Moore is holding 
up is a hand." 


There are two connected principles about psychological concepts that are fundamen- 
tal for Malcolm. One is: Only of a living human being and what resembles (behaves like) 
a living human being can one say: it has sensations; it sees; is blind; hears; is deaf; 
is conscious or unconscious . . . thinks (see Wittgenstein, Philosophical Investigations 
§282, §360). We ascribe mental properties to others on the basis of observable 
behavioral criteria that are non-contingently connected to the concepts of those 
properties; it is part of having the concepts of the properties to know behavioral 



criteria that justify ascribing the properties in the right circumstances. The second is: 
It is not on the basis of any criteria that we ascribe mental properties (current, 
conscious ones) to ourselves; our self-ascriptions are analogous to, and in some cases 
simply replace, natural manifestations of mental states, such as the expression of pain 
in crying or moaning; they serve others as criteria for ascribing the properties to us. 
"First person utterances, and their second and third person counterparts," he says, "are 
linked in meaning by virtue of being tied, in different ways, to the same behavioral 
criteria" (1971: 91). Cartesian philosophy of mind, which he sometimes calls 
"introspectionism," runs afoul of the first principle. Behaviorism as a philosophy of 
mind runs afoul of the second. 

Against the thesis that mental states or processes are identical with brain states or 
processes Malcolm marshals several different arguments. In "Scientific Materialism and 
the Identity Theory" (1964) he argues against J. J. C. Smart's claim that a sudden 
thought is contingently identical with a brain process as follows: we attach no meaning 
to determining the bodily location of a thought; so, if x is identical with y only if x and 
y occur at the same place and time, and the identity is contingent, then there can be 
no way of establishing that this same location condition is satisfied. (In the same paper 
he remarks that the senselessness of the supposition that a separated brain could have 
thoughts or sensations "seems so obvious that I find it hard to take it seriously" (p. 
124).) In "Functionalism in Philosophy of Psychology" (1980) he imagines Mr. A 
saying to his wife, "Are you always on time?" and argues that one knows that Mr. A 
meant his utterance sarcastically and not admiringly only by knowing something of 
the previous course of their lives together, so that there is no way in which the pres- 
ence of the one intention rather than the other can be accounted for by some story 
about neural firings or electric potentials within Mr. A at the time of his utterance. In 
Consciousness and Causality (1984) he argues that mental states without genuine dura- 
tion (abilities, dispositions, intentions, beliefs) cannot be identical with brain states 
which do have genuine duration, and he argues that, since having an intention with a 
certain content entails having the concepts required to understand that content, it is 
impossible to identify the intention with a brain state, because possession of those con- 
cepts would, presumably be identified with other brain states and it is only contingent 
that they occur in the same brain. 

In his "The Conceivability of Mechanism" (1968), which has been much cited in 
subsequent discussions of mental causation, Malcolm argues that a completely mecha- 
nistic explanation of a piece of human behavior - one entirely in terms of physical 
states and processes in the organism - is incompatible with any intentional or purpo- 
sive explanation of it. He finds untenable both of the two ways he sees of trying to main- 
tain their compatibility: maintaining that intentional concepts can be defined in terms 
of non-intentionally specified behavioral dispositions and maintaining that intentional 
states or events are contingently identical with neurophysiological states or events. If 
all human behavior had sufficient mechanistic causes then, he argues, human beings 
would have no intentions or desires. And, he observes, there would therefore be 
a pragmatic paradox in anyone's asserting that all human behavior is mechanistically 
explicable: since the asserter's utterance could count as an assertion only if he has 
certain intentions about it, his asserting this would constitute a counterexample to 
what he asserts. 



Inspired by a remark of Wittgenstein's about dreaming, Malcolm notoriously 
argued, in a paper "Dreaming and Skepticism" (1956) and a book Dreaming (1959), 
that dreams cannot take place during sound sleep, in the sense of occurring at definite 
times and having definite durations. He infers this (and the stronger conclusion that 
there can be no mental activity during sound sleep) from the premise that the concept 
of sound sleep precludes the subject's manifesting any mental activity while sound 
asleep. He says that, if we found a correlation between some physiological process 
during sleep and reports on awaking of dreams and used that as a basis for locating 
dreams in objective time, that would be to adopt a different use of "dreaming" than we 
now have, a new meaning for the term. 'As things are," he says, "the notions of dura- 
tion and time of occurrence have no application in ordinary discourse to dreams. In 
this sense, a dream is not an 'occurrence' and therefore not an occurrence during sleep" 
(1956: 30). 

Malcolm made a significant contribution to the study of Descartes's philosophy of 
mind in two papers, "Descartes' Proof that His Essence is Thinking" (1965) and 
"Descartes' Proof that He is Essentially a Non-material Thing" (1975). The first con- 
jectures that Descartes argues as follows: "x is my essence if it is the case that (a) if I 
am aware of x then (necessarily) I am aware of myself, and (b) if I am aware of myself 
then (necessarily) I am aware of x. Thinking satisfies these conditions. Ergo, thinking 
is my essence" (1977b: 32). This argument, Malcolm suggests, could be Descartes's 
reason for thinking that he has a clear and distinct idea of himself as a thing with no 
corporeal characteristics. Malcolm's criticism of the argument is that, although (a) and 
(b) are true when "thinking" is substituted for x, this is not because of any necessary 
connection between myself and thinking. He says: "(a) is true solely because the state- 
ment 'I am not aware of myself is self-defeating . . . (b) is true because the awareness 
of anything is thinking, and also because of Descartes' doctrine that one cannot think 
without being aware of thinking" (1977b: 36). 

In the second paper, Malcolm, responding to a suggestion from Robert Jaeger, finds 
textual support in Descartes for the following argument: "I think I am breathing entails 
I exist. I think I am breathing does not entail I have a body. Therefore, I exist does not 
entail I have a body." Malcolm rejects the second premise. It is, he says, conceptually 
impossible for me to exist without ever having had a body, or for minds to exist without 
there ever having been bodies, because the primary use of "He thinks he is breathing" 
presupposes behavioral criteria of its truth (and secondary uses in speaking of ghosts 
or disembodied deities could not exist without primary uses). He points out that 
Descartes could be hoist with his own petard here, for I am breathing entails I exist but 
does not entail I am thinking. 

In "Thoughtless Brutes" (1972 presidential address to the Eastern Division of the 
American Philosophical Association) Malcolm argues that the reason Descartes claims 
that animals do not have "real" sensations is that he insists that "when we mean by 
'sensation' something other than mere physiological processes, then sensation [has] 
propositional content" and he thinks, rightly, that propositional representations do 
not occur in the "lower" animals. Malcolm comments, "When we see the enormity of 
[Descartes's] exaggeration of the propositional in human life, our unwillingness to 
ascribe propositional thinking to animals ought no longer to make us refuse to attribute 
to them a panoply of forms of feeling, of perception, of realization, of recognition, that 



are, more often than not, nonpropositional in the human case" (1972: 53). He adds, 
"We need to avoid identifying thoughts with their linguistic expression. At the same 
time we should reject the suggestion that it is possible that language-less creatures 
should have thoughts . . . [F]or it is meaningful to suppose that a person might have 
had a thought to which he gave no expression, only because this person speaks or spoke 
a language in which there is an institution of testifying to previously unexpressed 
thoughts" (p. 55). 


Malcolm's work on memory is found in his "Three Lectures on Memory" (1963b) and 
a book, Memory and Mind (1977a). In the first lecture, "Memory and the Past," he 
argues against Russell that the hypothesis that the world began five minutes ago com- 
plete with misleading records, delusory memories, etc., is not "logically tenable." His 
main argument is that a linguistic community can be said to have mastered the past 
tense, and therefore make past tense statements and have past tense beliefs, only if not 
all of their past tense statements are false. He also asserts that, if our apparent memo- 
ries largely agree with each other and with the records then the apparent memories 
would be verified as true, and "if the apparent memories were verified it would not be 
intelligible to hold that, nevertheless, the past they describe may not have existed" 
(1963a: 199). 

In the second lecture, "Three Forms of Memory," he distinguishes factual memory 
(remembering that p), personal memory (remembering something one previously 
perceived or experienced), and perceptual memory (personally remembering some- 
thing by forming a mental image of it). He says that while a personal or perceptual 
memory always entails some factual memory, there can be a factual memory that 
does not entail any perceptual or personal memory (contrary to Russell and others). 
There could, he says, be a person who lacked perceptual memory altogether but 
had more or less normal factual and personal memories, but there could not be a 
creature we would recognize as a human being who altogether lacked factual or 
personal memory. 

In the third lecture, "A Definition of Factual Memory," he suggests the following 
definition: "A person, B, remembers thatp from a time, t, if and only if B knows that 
p, and B knew that p at t, and if B had not known at t that p he would not now know 
that p" (1963a: 236). Concerning the third, counterfactual conjunct here, he says, 
"Whether or not it makes sense to postulate a specific brain-state or neural process per- 
sisting between the previous and the present knowledge that p, such a postulation is 
obviously not required by an analysis of the concept of remembering," and guesses 
"that our strong desire for a mechanism of memory arises from an abhorrence of the 
notion of action at a distance-in-time" (1963a: 23 7-8). 

In the book he maintains that it is an error to think that the causal ingredient in 
memory requires the assumption either of a temporally continuous chain of causation 
or of causal laws. He argues against the idea that there must be a representation in 
remembering and the idea that there must be a structural isomorphism between an 
occurrent memory, what is remembered, and an intervening brain state or process: 
what one remembers of a remembered experience could not be enumerated in a closed 



set of items of the sort needed to make out an isomorphism, and it would be impos- 
sible to devise a key of isomorphism that could provide any reasonable prospect for the 
discovery of a one-to-one correlation of component elements between any mental state 
and any neural state. 

Philosophy of religion 

Malcolm's paper "Anselm's Ontological Arguments" (1960) provoked considerable 
discussion. In it he says that Anselm put forward two different ontological proofs 
of the existence of God. The first, in Proslogion 2, uses the principle that a thing is 
greater if it exists than if it does not exist. The second, in Proslogion 3 , uses the differ- 
ent principle that a thing is greater if it necessarily exists than if it does not necessar- 
ily exist. The first is fallacious because it is an error to regard existence as a property 
of things that have contingent existence, but it does not follow that it is an error 
to regard necessary existence as a property of God and as a perfection. A short 
summary of the second proof: If God exists, His existence is necessary; thus God's 
existence is either necessary or impossible; assuming that the concept of God is not 
self-contradictory or in some way logically absurd, it follows that He necessarily 
exists. Malcolm remarks, "I should think there is no more a presumption that 
[the concept of God] is self-contradictory than is the concept of seeing a material 
thing. Both concepts have a place in the thinking and the lives of human beings" 
(1963a: 160). 

Bibliography of Malcolm's work 

A complete list of Malcolm's articles published through 1981 may be found in Carl Ginet and 
Sydney Shoemaker (eds.) Knowledge and Mind (Oxford: Oxford University Press, 1983). The 
collection Wittgensteinian Themes (1995) contains fourteen of his essays written in the last 
twelve years of his life. 

1942: "Certainty and Empirical Statements," Mind 51, pp. 18-46. 

1949: "Defending Common Sense," Philosophical Review 58, pp. 201-21. 

1950: "The Verification Argument," in Philosophical Analysis, ed. M. Black, Ithaca, NY: Cornell 

University Press. (Reprinted with revisions and additional footnotes in Malcolm 1963a.) 
1951: "Philosophy for Philosophers" (intended title: "Philosophy and Ordinary Language"), 

Philosophical Review 60, pp. 329-40. 
1956: "Dreaming and Skepticism," Philosophical Review 65, pp. 14-3 7. 
1959: Dreaming, London: Routledge and Kegan Paul. 
1960: 'Anselm's Ontological Arguments," Philosophical Review 69, pp. 41-60. (Reprinted with 

new footnotes in Malcolm 1963a.) 
1963a: Knowledge and Certainty, Englewood Cliffs, NJ: Prentice-Hall. 
1963b: "Three Lectures on Memory," ("Memory and the Past," "Three Forms of Memory," and 

"A Definition of Factual Memory"), in Malcolm 1963a. ("Memory and the Past" first published 

in The Monist 45 (1962), pp. 247-66.) 
1964: "Scientific Materialism and the Identity Theory," Dialogue 3, pp. 115-25. 
1965: "Descartes' Proof that His Essence is Thinking," Philosophical Review 74, pp. 315-38. 

(Reprinted in Malcolm 1977b.) 
1968: "The Conceivability of Mechanism," Philosophical Review 77, pp. 45-72. 



1971: Problems of Mind, New York: Harper and Row. 

1972: "Thoughtless Brutes," Presidential address, Proceedings of the American Philosophical 

Association 46, pp. 5-20. (Reprinted in Malcolm 199 7b.) 
1975: "Descartes' Proof that He is Essentially a Non-material Thing," Philosophy Forum 14 

(19 75). (Reprinted in Malcolm 19 77b.) 
19 76: "Moore and Wittgenstein on the Sense of T know'," in Essays in Honour ofG. H. von Wright, 

in Acta Philosophica Pennica ed. Jaakko Hintikka, 28, 1-3, pp. 216-40. 
1977a: Memory and Mind, Ithaca, NY: Cornell University Press. 
1977b: Thought and Knowledge, Ithaca, NY: Cornell University Press. 
1980: "Functionalism in Philosophy of Psychology," Proceedings of the Aristotelian Society, new 

series 80, pp. 211-29. 
1984: Consciousness and Causality: A Debate on the Nature of Mind with D. M. Armstrong, Oxford: 

Blackwell Publishers. 
1986: Wittgenstein: Nothing is Hidden, Oxford: Blackwell Publishers. 
1994: Wittgenstein: A Religious Point of View?, ed. with a response by P. Winch, Ithaca, NY: Cornell 

University Press. 
1995: Wittgensteinian Themes: Essays 1978-1989, ed. G. Henrik von Wright, Ithaca, NY: Cornell 

University Press. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

Wilfrid Sellars (1912-1989) 


Life and work 

Had Wilfrid Stalker Sellars never written an original philosophical word, his accom- 
plishments as an editor would likely be sufficient to earn him a place of honor in the 
history of postwar analytic philosophy. In 1950, he and Herbert Feigl co-founded 
Philosophical Studies, the first scholarly journal explicitly devoted to analytic philosophy, 
which they edited jointly until 1971 and Sellars then edited alone for three more years. 
A year earlier, Feigl and Sellars had already published a seminal anthology, Readings in 
Philosophical Analysis; Readings in Ethical Theory, co-edited by Sellars and John Hospers, 
followed three years later. The "philosophical analysis" represented in these volumes 
had been transplanted from its origins and early development at Cambridge and Oxford 
and enriched by generous cross-fertilization from the logical empiricism of an expatri- 
ate Vienna Circle, most notably by the work of Rudolf Carnap, and indigenous strains 
of pragmatism, critical realism, and evolutionary naturalism. From such seeds, the 
"analytic" style of philosophizing and its agenda of problems grew to dominate 
American academic philosophy, definitively changing its intellectual landscape. The 
"logico-linguistic turn" became the new methodological center of philosophical 
inquiry, and regional philosophies of logic, language, mind, and science first joined and 
then gradually began to supplant more broadly-conceived traditional metaphysical and 
epistemological studies, while normative ethical inquiries gave ground to issues in 
metaethics and moral psychology. 

But Wilfrid Sellars in fact wrote many an original and important philosophical word, 
and so not only helped to stimulate the growth of postwar analytic philosophy, but also 
became one of its most distinguished and influential practitioners. His academic tra- 
jectory took him from studies at the University of Michigan and, as a Rhodes Scholar, 
at Oxford University, to faculty positions at the University of Iowa, the University of 
Minnesota, where he served as Chair during the mid-1950s, and Yale University, before 
he became University Professor of Philosophy and Research Professor of the Philosophy 
of Science at the University of Pittsburgh, a position which he held from 1963 until 
his death in 1989. His intellectual trajectory meanwhile carried him from an early 
period, during which he worked out his fundamental philosophical ideas in over two 
dozen dialectically-challenging essays, through a middle period characterized by the 



development and exposition of a systematic philosophical vision of remarkable scope 
and depth, into a late period of consolidation, refinement, and deepening of mature 
theses and insights that were simultaneously coming to be more fully appreciated and 
explicitly appropriated by a new philosophical generation. 

The critique of givenness 

Sellars's revolutionary 1956 essay "Empiricism and the Philosophy of Mind," immedi- 
ately acknowledged as a contemporary classic, marks the beginning of his exception- 
ally productive and influential middle period. (This can be treated, somewhat arbitrarily, 
as culminating in 1972 with the publication of his 1970 American Philosophical 
Association Eastern Division Presidential Address on the Kantian text, "this I or he or 
it (the thing) which thinks.") The central theme of "Empiricism and the Philosophy of 
Mind" is a thoroughgoing and general critique of what Sellars famously called the 
"myth of the given," a perennial and polymorphic philosophical motif manifested inter 
alia in the idea, characteristic of classical sense-datum theory, that empirical knowl- 
edge rests on a foundation of "immediate awarenesses" and on the assumption that the 
"privacy" of the mental and one's "privileged access" to one's own mental states are 
primitive features of experience, logically and epistemologically prior to all intersub- 
jective concepts pertaining to inner episodes. 

Sellars criticized sense-datist and other traditional epistemic foundationalisms for 
failing properly to distinguish non-conceptual states of sensing from conceptually 
structured perceptual takings. Perceiving always involves taking something sensorily 
present to be this or that, and so, as Kant recognized, has a judgmental form which 
mobilizes and applies, correctly or incorrectly, descriptive and classificatory concepts. 
Perception properly so-called is consequently a normative business, and the ability to 
engage in it requires more than reliable differential dispositions to respond to sensory 
stimuli. "The essential point," Sellars wrote, "is that in characterizing an episode or a 
state as that of knowing, we are not giving an empirical description of that episode or 
state; we are placing it in the logical space of reasons, of justifying and being able to 
justify what one says" (1963a: 169). A perceptual judgment may consequently be direct 
or immediate in the sense of being a unmediated response to stimulation, i.e. not itself 
inferred from other judgments, but the epistemic authority of such judgments depends 
upon their being appropriately caught up in the intersubjective game of having and 
giving reasons, and so cannot be independent of their inferential relationships to other 
judgments. Sellars thus notoriously advocated a strong epistemic internalism, according 
to which "observational knowledge of any particular fact . . . presupposes that one 
knows general facts of the form Xis a reliable symptom of Y" (1963a: 168). 

Epistemic authority 

Consistent with his strong internalism, Sellars interpreted even first-person epistemic 
authority with respect to the sensory aspects of one's own experience as built on and 
presupposing an intersubjective status for sensory concepts perse. Correlatively, he deci- 
sively rejected the idea that sensory consciousness supplies a form of empirical knowl- 
edge that (1) is immediate (i.e. non-inferential); (2) presupposes no knowledge of other 



matters of fact, particular or general; and (3) constitutes the ultimate court of appeals 
for all factual claims (1963a: 164). Although a person can justifiably believe an empiri- 
cal truth without having inferred it from other propositions, no empirical beliefs are 
self-justifying, self-warranting, or self-authenticating. Instead, Sellars argued, "to say 
that someone directly knows that-p is to say that his right to the conviction that-p essen- 
tially involves the fact that the [thought] that-p occurred to the knower in a specific 
way" (p. 188). 

The epistemic authority of a non-inferential perceptual belief can be traced to the 
fact that, in the course of learning perceptual language, the believer has not only 
acquired propensities for the reliable use of the relevant concepts in perceptual situa- 
tions but also has come to know what is involved in learning to use perceptual sen- 
tences reliably in perceptual contexts. Thus, assuming that he has mastered the use of 
the relevant words in suitable perceptual situations, a person who candidly and spon- 
taneously thinks-out-loud "Lo! Here is a red apple" - Sellars's customary model of a 
perceptual taking - is justified in reasoning: 

I just thought-out-loud "Lo! Here is a red apple" (no countervailing conditions obtain); so, 
there is good reason to believe that there is a red apple in front of me. (19 75d: 341-2) 

This "trans-level" reasoning, as Sellars called it, does not have the original perceptual 
judgment as its conclusion, but is rather an inference from the character and context 
of the original non-inferential experience to the existence of a good reason for accept- 
ing it as veridical. 

Central to Sellars's thoroughgoing epistemic internalism, indeed, is his conviction 
that the reasonableness of accepting even first principles is a matter of the availability 
of good arguments warranting their acceptance. What is definitive of first principles 
(FP) is the unavailability of sound arguments in which they are derived as conclusions 
from still-more-basic premises, that is, arguments of the form: 

(Al) P 1( P a P n hFP. 

Here, too, Sellars appeals to the notion of a "trans-level" justificatory inference, point- 
ing out that the absence of good arguments of the form (Al) is entirely compatible with 
the existence of sound arguments of the form 

(A2) Pj, P, P m h It is reasonable to accept FP. 

Sellars interpreted the conclusion of (A2) as a claim to the effect that a particular 
course of epistemic conduct, accepting the principle FP, can be supported by adequate 
reasons. This suggested the in-principle availability of yet another argument, specifi- 
cally a piece of sound practical reasoning, whose conclusion expresses an intention to 
engage in such conduct: 

(A3) I shall achieve desirable epistemic end E. 

Achieving E implies accepting principles of kind K. 
The principle FP is of kind K. 
Therefore, I shall accept FP. 

On Sellars's view, such patterns of practical reasoning also govern the warranted 
acceptance of lawlike generalizations and theoretical systems. Adopting a systematic 



theoretical framework is ultimately justified by appeal to the epistemic end of "being 
able to give non-trivial explanatory accounts of established laws" (19 75c: 384), and 
adopting nomological claims that project the observed statistical frequency (including 
a frequency = 1) of some property in an open class to further unobserved finite samples 
of the class, is ultimately justified by appeal to the epistemic end of "being able to draw 
inferences concerning the composition with respect to a given property Y of unexam- 
ined finite samples ... of a kind X, in a way which also provides an explanatory account 
of the composition with respect to Y of the total examined sample, K, of X" (1975c: 
392). What is crucial is that these ends concern 

the realizing of a logically necessary condition of being in the framework of explanation 
and prediction, i.e. being able to draw inferences concerning the unknown and give 
explanatory accounts of the known. (1975c: 397) 

Sellars argued, in short, that inductive reasoning does not need to be vindicated, that 
is, shown to be truth-preserving, but is rather itself fundamentally a form of vindica- 
tive reasoning, justifying our engaging in determinate epistemic conducts. Its ends-in- 
view must consequently be of a sort that can be known to be realized or obtain. Unlike 
such Reichenbachian ends as being in possession of limit-frequency statements which 
are within a certain degree of approximation of the truth (where such limits exist), the 
ends of being in possession of explanatory laws and principles envisioned by Sellars 
satisfed that constraint. 


Sellars famously engaged the Cartesian picture of direct and incorrigible s?Z/-knowledge 
with his "myth of Jones," a story set in a fictional community whose members speak a 
hypothetical sophisticated "Rylean language." The fundamental descriptive vocabulary 
of this language pertains to public spatiotemporal objects, and while it includes logical 
operators, subjunctive conditionals, and even the fundamental resources of semantic 
discourse (enabling its speakers to say of their peers' utterances that they mean 
this or that, stand in various logical relations to one another, and are true or false), it 
nevertheless lacks any resources for speaking of inner episodes, whether thoughts or 

In this community, then, a genius, Jones, develops a theory according to which overt utter- 
ances are but the culmination of a process which begins with certain inner episodes. . . . 
[His] model for these episodes which initiate the events which culminate in overt verbal 
behavior is that of overt verbal behavior itself. (1963a: 186) 

Jones's theory earns its epistemic credentials by enabling him, and his fellow Ryleans 
who master it, successfully to explain and anticipate intelligent behavior conducted 
in silence, that is, unaccompanied by overt verbal episodes of the sort that we would 
recognize as expressing an agent's conduct-rationalizing beliefs and desires. 

The new idioms of Jones's theory, for example, "is thinking '. . .'," initially have 
a purely theoretical use, being ascribed on the basis of inferences from observable 



behavior in observable circumstances. But crucially, Sellars argued, what begin as 
purely theoretical terms can acquire a first-person reporting role. For it turns out to be 
possible for Jones to train his compatriots, in essence by a process of Skinnerian operant 
conditioning, to respond reliably and directly (i.e. non-inferentially) to the occurrence 
of such an "inner episode" with a judgment to the effect that it is occurring. That is, 
they can respond to one thought with a second (meta-)thought to the effect that they 
are thinking it; this matches the de facto phenomenology of first-person "privileged 
access" sufficiently to account for the Cartesian illusion of mental transparency. The 
Jonesean story thus shows how the essential intersubjectivity of language can be 
reconciled with the "privacy" of inner episodes. 

Scientific realism 

Sellars's novel appeal to forms of theoretical reasoning in his myth of Jones reflected 
his broader philosophical concern with the nature, structure, and role of theories in 
the natural sciences. On the received, positivist, view explanation was identified with 
derivation. Singular matters of empirical fact were to be explained by deriving descrip- 
tions of them from ("inductive") empirical generalizations (along with appropriate 
statements of initial conditions), and these "empirical laws" in turn were to be 
explained by deriving them from theoretical postulates and correspondence rules. On 
the positivist view, in consequence, theories (e.g. microtheories) explain observational 
matters of fact only indirectly, by implying the (observation-language) generalizations 
that explain them directly. 

Sellars, in contrast, argued that this "levels picture" of theories was fundamentally 
misleading. Theories do not explain laws by entailing them. Rather, "theories explain 
laws by explaining why the objects of the domain in question obey the laws that they 
do to the extent that they do" (1963c: 123). 

[That is,] they explain why individual objects of various kinds and in various circum- 
stances in the observation framework behave in those ways in which it has been induc- 
tively established that they do behave. Roughly, it is because a gas is ... a cloud of 
molecules which are behaving in certain theoretically defined ways, that it obeys the 
empirical Boyle-Charles Law. (1963c: 121) 

On Sellars's view, then, "theoretical entities" are not merely convenient fictions, 
enabling us to abbreviate complicated and unwieldy stories about entities that we 
have good, (observational) reasons to believe actually exist. Theoretical entities are 
rather those entities we justifiedly believe to exist for good and sufficient theoretical 

Sellars thus advocated a robustly realistic epistemology of scientific inquiry and, cor- 
relative ly an understanding of its ultimate outcomes as onto logically definitive: "In the 
dimension of describing and explaining the world, science is the measure of all things, 
of what is that it is, and of what is not that it is not" (1963a: 173). 

(This is his "scientia mensura.") Scientific theories indeed explanatorily "save the 
appearances," but they do so precisely by describing the reality of which the appear- 
ances are appearances. This robust realism, combined with a thoroughgoing 



naturalism, in fact set Sellars's metaphilosophical agenda for postwar analytic 
philosophy per se. 

Metaphilosophical views 

Sellars saw contemporary philosophy as confronted by two "images," each of which 
purported to be a complete picture of man-in-the-world, which need to be unified into 
a single synoptic vision. The "manifest image" is the conception of the world and the 
place of persons in it that has descended from the great speculative systems of ancient 
Greece, through the dialectics of a "perennial philosophy," to the dimensions of con- 
temporary Anglo-American thought that emphasize "ordinary usage" and "common 
sense." Its primary objects are persons, beings who, inter alia, reflectively conceive of 
themselves as being in the world both as sentient perceivers and cognitive knowers of 
it, and as agents capable of affecting it through deliberate and rational elective conduct. 
In contrast, the "scientific image" is the complex new understanding of man-in-the- 
world that is still in the process of emerging from the fruits of theoretical reasoning, in 
particular, from the processes of postulational theory construction. Although this 
image is "methodologically dependent on the world of sophisticated common sense," 
Sellars argues that 

it purports to be a complete image, i.e. to define a framework which could be the whole 
truth about that which belongs to the image. Thus although methodologically a develop- 
ment within the manifest image, the scientific image presents itself as a rival image. From 
its point of view the manifest image on which it rests is an "inadequate" but pragmatically 
useful likeness of a reality which first finds its adequate (in principle) likeness in the 
scientific image. (1963c: 57) 

The leading challenge for contemporary philosophy, he concluded, is to show how the 
inevitable tensions between these two images can be resolved by a "stereoscopic under- 
standing" in which they come to be "fused" into a single synoptic vision of man-in- 
the-world. Sellars's philosophy is usefully viewed as a fuller articulation of this con- 
frontation of the two images and a detailed dialectical engagement with the philo- 
sophical agenda to which it gives rise: that places be found within the context of a 
thoroughgoing naturalism for the intentional contents of language and thought, for 
the normative dimensions of knowledge and action, and for the sensuous contents of 
perception and imagination. Consonant with such a naturalism, the sought synoptic 
story must find a place for mind without assigning an independent, autonomous, and 
irreducible, ontological status to intentional states or entities, and it must eschew any 
ontological view of abstracta as real that fails to deliver an adequate account of their 
role within the causal order, broadly construed. 

Semantic meaning 

The centerpiece of Sellars's response to both of these naturalistic challenges was a 
sophisticated theory of conceptual roles, concretely instantiated in the conducts of lin- 
guistic communities and socially transmitted by modes of cultural inheritance. At the 



heart of this theory was an increasingly refined account of meaning as functional 
classification, more precisely, of the "meaning" idiom as, in the first instance, a context 
of translation in terms of which structurally distinct "natural-linguistic objects" (e.g. 
utterings or inscribings) are classified in terms of their roles or functions vis-a-vis the 
organized behavioral economies of families of speaking organisms, fn short, Sellars 
interpreted "means" as a specialized form of the copula, tailored to metalinguistic con- 
texts, according to which the right side of the superficially relational form "- means 
..." is also properly understood as mentioning or exhibiting a linguistic item. 

Ordinary quotation, argued Sellars, suffers from a systematic ambiguity regarding 
the criteria according to which linguistic tokens are correctly classifiable as belonging 
to this or that linguistic type. He therefore introduced the straightforward device of 
two separate styles of quotation marks - star-quotes and dot-quotes - to differentiate 
between two different ways of sorting and individuating lexical items. Star-quotes form 
common nouns that are true of items belonging to the spatiotemporal causal order 
("tokens") which are appropriately structurally isomorphic to the tokens exhibited 
between them, while dot-quotes form common nouns true of tokens that, in some spec- 
ified language, are appropriately functionally isomorphic to (i.e. play the role or do the 
job performed by) the tokens exhibited between them in our language. Sellars then 
proposed to transcribe such semantic claims as 

(Is) (In French) "rouge" means red, 


(2 s) (In German) "Es regnet" means it is raining, 

by the more perspicuous formulations 

(1*) (In the French linguistic community) *rouge*s are 'red's 


(2*) (In the German linguistic community) *Es regnet*s are 'it is raining's. 

Roles and rules 

To sort and classify descriptively individuated families of natural items in terms of their 
linguistic jobs, roles, or functions, is to sort and classify them normatively. Sellars's 
overall philosophical agenda thus required a complementary thoroughly naturalistic 
account of the normative dimension of language, and he indeed offered one, basing it 
on the notion of what he called pattern-governed behavior. The basic concept of pattern- 
governed behavior is 

the concept of behavior which exhibits a pattern, not because it is brought about by the 
intention that it exhibit this pattern, but because the propensity to emit behavior of the 
pattern has been selectively reinforced, and the propensity to emit behavior which does not 
conform to this pattern selectively extinguished. (19 74: 423) 

Pattern-governed behavior can arise from processes of natural selection on an evolu- 
tionary time-scale as a characteristic of a species, for example, the dance of the bees, 



but it can also be developed in individuals, "trainees," by deliberate and purposive selec- 
tion on the part of other individuals, the "trainers." In this connection, Sellars intro- 
duced a distinction between "rules of action" and "rules of criticism." 

Rules of action specify what someone ought to do, for example, "Ceteris paribus, one 
ought to say such and such if in circumstances C." They can be efficacious in guiding 
linguistic activity only to the extent that their subjects already possess the relevant con- 
cepts, such as concepts of "saying such-and-such," of "being in circumstances C," and, 
indeed, of obeying a rule (i.e. doing something because it is enjoined by a rule). Rules 
of criticism, in contrast, specify what ought to be the case, for example, "Westminster 
clock chimes ought to strike on the quarter hour" (1975d: 95). The items whose 
performances may legitimately be appraised according to such rules, however, need 
not themselves have the concept of a rule nor, indeed, even be capable of having any 
concepts at all. 

Thus trainers can be understood in the first instance as acting in accordance with 
rules of conduct whose authority derives from rules of criticism, that is, as aiming at 
bringing about the pattern-governed behaviors which their trainees' conduct ought to 
manifest: "Patterned-behavior of such and such a kind ought to be exhibited by 
trainees, hence we, the trainers, ought to do this and that, as likely to bring it about 
that it is exhibited" (1974: 423). 

If training is successful, then, in consequence of the conducts of trainers under the 
guidance of such rules of action, the behavior of a language-learner can come to 
conform to the relevant rules of criticism without his, in any other sense, grasping them 

[The] members of a linguistic community are first language learners and only potentially 
"people," but subsequently language teachers possessed of the rich conceptual framework 
this implies. They start out being the subject matter of the ought-to-be's and graduate to 
the status of agent subjects of the ought-to-do's. (19 75d: 100) 

The modes of pattern-governed behavior relevant to semantic meaning and, corre- 
spondingly, the relevant families of rules of criticism, Sellars proposed, are threefold: 

1 Language Entry Transitions: It ought to be the case that speakers respond to objects 
in perceptual situations and to certain states of themselves with appropriate 
linguistic -conceptual activity. 

2 Intra-linguistic Moves: It ought to be the case that speakers' linguistic-conceptual 
episodes tend to occur in patterns of valid inference (theoretical and practical, 
formal and material), and tend not to occur in patterns which violate logical 

3 Language Departure Transitions: It ought to be the case that speakers respond to 
such first-person linguistic -conceptual episodes as "I shall now raise my hand" with 
an upward motion of the hand, etc. (Cf. 1974: 423-4) 

These transitions are respectively the core elements of perceptual takings, inferences, 
and volitions, and, although they are acts {qua both actualities and actualizations of 
behavioral dispositions), Sellars insisted that they are not themselves actions. They are 
acquired as, and remain, pattern-governed activities, but form the basis of linguistic 



actions proper as "the trainee acquires not only the repertoire of pattern-governed lin- 
guistic behavior which is language about non-linguistic items, but also that extended 
repertoire which is language about linguistic as well as non-linguistic items" and thus 
becomes "able to classify items in linguistic kinds, and to engage in theoretical and 
practical reasoning about his linguistic behavior" (1974: 425). 

Linguistic roles or functions are individuated in terms of the structure of positive 
and negative uniformities generated in the natural order by these pattern-governed 
activities of perception, inference, and volition. Analyzing sameness of linguistic role 
as sameness of place in the complex relational structure generated by conducts that 
have been causally shaped in these ways by systems of espoused linguistic norms 
equipped Sellars with a functional conception of semantics. This conception neither 
presupposed nor unavoidably led back into the domains of either ontological abstracta 
or irreducibly intentional mental entities. 

The intentionality of thought 

Instead, Sellars's alternative account of the distinctive intentionality of thought was 
itself drawn in terms of the forms and functions of natural linguistic items, modeled by 
what he came to call "verbal behaviorism" (VB): 

According to VB, thinking "that-p," where this means "having the thought occur to one 
that-p," has as its primary sense [an event of] saying "p"; and a secondary sense in which 
it stands for a short term proximate propensity [disposition] to say "p." (1974: 419) 

The roots of Sellars's mature verbal behaviorism reach back to the myth of Jones in 
"Empiricism and the Philosophy of Mind." The thought-episodes postulated by Jones 
on the model of overt verbal behavior are introduced by a purely functional analogy. 
The concept of an occurrent thought is not that of something encountered propria 
persona but rather that of a causally-mediating logico-semantic role-player, whose 
determinate ontological character is initially left entirely open. 

Since on Sellars's account the concept of a thought is fundamentally the concept of 
a functional kind, no ontological tension is generated by the identification of items 
belonging to that functional kind with states and episodes of an organism's central 
nervous system. The manifest image's conception of persons as thinkers can conse- 
quently fuse smoothly with the scientific image's conception of persons as complex 
material organisms having a determinate physiological and neurological structure. 
Sellars's conviction that the fundamental characteristic of semantic discourse is its ine- 
liminable appeal to functional considerations, and his correlative pioneering analyses 
of the intentional categories of the mental in terms of epistemologically theoretical 
transpositions of the semantic categories of public language earn him the title of the 
first functionalist in contemporary analytic philosophy of mind. This functionalist view 
is one whose implications and influence have not yet begun to be exhausted. 

Categorial ontology 

The parallels between semantic discourse and the classical ontological idioms of 
Platonistic discourse, ostensibly designating abstract entities, have not gone unnoticed. 



Consistent with his global commitment to naturalism, Sellars exploited these parallels 
to construct his own unique variant of linguistic nominalism, a view according to 

the abstract entities which are the subject matter of the contemporary debate between 
platonistic and anti-platonistic philosophers - qualities, relations, classes, propositions, 
and the like - are linguistic entities. (1967a: 229) 

In first approximation, Sellars proposed to analyze the ostensibly relational ontological 

(lo) (The French word) "rouge" stands for redness, 


(2o) (The German sentence) "Es regnet" expresses the proposition that it is raining, 

precisely as he had analyzed the corresponding semantic claims (Is) and (2s). These 
two will be analyzed as (1*) and (2*), claims that specify the functional roles of deter- 
minate families of structurally-individuated tokens. This strategy of understanding 
traditional ontological discourse as classificatory discourse within a functional meta- 
language transposed into the "material mode of speech" had been pioneered by Carnap 
(see CARNAP). But unlike Carnap, Sellars refused to identify the formally definable con- 
structs of a "pure" syntax or semantics with the corresponding syntactical and seman- 
tical terms in everyday, pre-philosophical usage. Such a facile interpretation of the 
relationship between "pure" and "descriptive" syntactic and semantic discourses, he 
argued, fails to do proper justice to the essential normative dimension of the latter. 
Thus, while Sellars was prepared to interpret such paradigmatic ontological terms as 
"universal," "individual," "kind," "quality," "proposition," and "fact" by appealing to 
syntactic and semantic counterparts (e.g. "predicate," "singular term," "common 
noun," "monadic predicate," "sentence," and "true sentence") he insisted that these 
syntactic and semantic terms 

have a conceptual role which is no more reducible to [non-syntactical and] non- 
semantical roles than the role of prescriptive terms is reducible to non-prescriptive roles. 
. . . [The] empirical (in the broad sense) character of statements in descriptive (historical) 
[syntax and] semantics does not entail that [syntactical and] semantical concepts, 
properly so called, are descriptive. (1975a: 459) 

Like Sellars's account of the distinctive intentionality of thought, then, his account of 
discourse ostensibly about the entities and categories of classical ontology is also drawn 
in terms of the forms and functions of natural linguistic items. "Abstract entities," too, 
consequently constitute no obstacle to the sought fusion of the manifest and scientific 


Surprisingly it is when Sellars turns from intentional thought and ontological abstracta 
to sensations that significant complications to his synoptic project first come into view. 



Like Kant, Sellars rejected the Cartesian picture of a sensory-cognitive continuum. 
The "of-ness" of a sensation - e.g. its being of a red triangle or of a sharp shooting 
pain - he insisted, is not the intentional "of-ness" ("aboutness") of a thought. "The 
'rawness' of 'raw feels'," he wrote, "is their non-conceptual character" (1967c: 3 76). 
Consequently, although Sellars's epistemological story about sensations also begins with 
a strategic appeal to the unique epistemic status of postulated theoretical entities, his 
account of the ontology of sensations diverges significantly from his semantic and 
functionalist account of intentional thoughts. 

In a final episode of the Myth of Jones the hero . . . postulates a class of inner - theoreti- 
cal - episodes which he calls, say, impressions, and which are the end results of the 
impingement of physical objects and processes on various parts of the body. (1963a: 191) 

Jones postulates impressions as elements of an explanatory account of the occurrence 
in various circumstances of perceptual cognitions with determinate semantic contents. 
In this case, however, the model for Jones's theory is not functionally-individuated fam- 
ilies of sentence tokens, but rather "a domain of 'inner replicas' which, when brought 
about in standard conditions share the perceptible characteristics of their physical 
sources" (1963a: 191). Although the entities of this model are particulars, the entities 
introduced by Jones's theory are not particulars but rather non-conceptual (non- 
intentional) states of a perceiving subject. Thus, while talk of the "of-ness" of sensa- 
tions, like that of the "of-ness" of thoughts, is fundamentally classificatory, the classi- 
fication of sensations is not functional, but rather based on analogies that are 
initially extrinsic and causal, and ultimately intrinsic and contentive. 

In the first instance, then, the concept of a person's having an of-a-red-triangle sen- 
sation - an adjectival idiom contrived to highlight the classificatory role of "of-ness" - 
or the concept of a person's sensing [red triangle] s ly - an adverbial idiom contrived to 
reflect the fact that "sensation" is a "verbal noun" - is the concept of her being in the 
sort of state that is brought about in normal perceivers in standard conditions by the 
action of red triangular objects on the eyes. The point of the model of "inner replicas," 
however, is to insist that such states can discharge their explanatory jobs in relation to 
cognitive perceptual takings (especially non-veridical perceptual judgments) only if 
they are conceived as having themselves determinate intrinsic characters and, in par- 
ticular, as resembling and differing from other sensory states in a manner formally anal- 
ogous to the way in which objects of the "replica" model (e.g. "wafers" of various colors 
and shapes) are conceived to resemble and differ from one another. 

Sellars proceeded to develop this core account of sensations in two different direc- 
tions, in consequence of which it has come to be regarded as one of the most difficult 
and controversial aspects of his philosophy. The first line of development turned on his 
conclusion that the fundamental concept pertaining to color within the manifest image 
is the concept of a kind of stuff. It is the concept of a quantum of red in space, an 
expanse or volume consisting of red, and is a basic concept in the sense that there is 
"no . . . determinate category prior to the concept of red as a physical stuff, as a matter 
for individuated physical things" (1981, I: 84). When dialectical pressures generate 
worries about the ontological status of the red which one ostensibly sees when it is not 
a constituent redness of a physical object, however, no alternative category can simply 



be "read off" from an introspective scrutiny of color quanta. The idea that a person is 
always also aware of the actual categorial status of the items that he encounters in per- 
ception or introspection, Sellars suggests, is only a particularly pernicious form of the 
myth of the given. 

All that is available is such transcendentals as "actual," "something" and "somehow." The 
red is something actual which is somehow a portion of red stuff, somehow the sort of item 
which is suited to be part of the content of a physical object, but which ... is not, in point 
of fact, a portion of physical stuff. (1981,1: 90) 

It then becomes the job of analogical thinking to construct new categorial forms of 
concepts pertaining to color. 

The first complication of Sellars's account of sensation resulted from his conviction 
that, in the case of sensations, Jones's theory takes this interpretive form. It does not 
introduce new domains of entities, but rather new forms of concepts. 

[Jones's] theory of sense impressions . . . reinterprets the categorial status of the cubical 
volumes of pink of which we are perceptually aware. Conceived in the manifest image as, 
in standard cases, constituents of physical objects and in abnormal cases, as somehow 
"unreal" or "illusory," they are recategorized as sensory states of the perceiver and 
assigned various explanatory roles in the theory of perception. (1981, III: 44) 

The crux of the Jonesean theory, in other words, is the thesis that the very color quanta 
of which we are perceptually aware as being in space are instead actually states of 
persons-qua-perceivers. It follows that, already within the manifest image, the onto- 
logical status ultimately accorded to sensory "content qualia" is in fact incompatible 
with their actually being instantiated in physical space. "[The] esse of cubes of pink is 
percipi or, to use a less ambiguous term, sentiri. Of course ... we are not perceptually 
aware of [them] as states of ourselves, though that is in point of fact what they are" 
(1981,111: 66). 

Absolute processes 

The second complication of Sellars's account of sensations then arose from his further 
conclusion that the scientific image's commitment to the idea that perceivers are 
complex systems of micro-physical particles constitutes a barrier to any straightfor- 
ward synoptic assimilation of this manifest image conception of sensory contents as 
states of perceiving subjects. On the one hand, Sellars observed that Jones's analogical 
treatment of sensory contents as states of perceivers formally preserves the "ultimate 
homogeneity" of those contents as originally conceived as space-filling stuffs. No 
defined states of a system or multiplicity of logical subjects, he argued, could continue 
to do so. 

On the other hand, Sellars contended, we cannot simply adopt a "reductive 
materialist" view according to which "what really goes on when a person senses 
a-cube-of-pinkly consists in [a certain] system of micro-physical particles being in a 
complex physical-2 state" (1981, III: 79), where "physical-2" states are those definable 



in terms of theoretical predicates necessary and sufficient to describe non-living 
matter. (To be "physical- 1," in contrast, is simply to belong in the space-time network.) 
For such reductive materialism amounts to the rejection of the idea that a (Jonesean- 
theoretical) state of, for example, sensing a-cube-of-pinkly is itself something actual in 
any categorial guise, and this fails properly to respect the philosophical demands 
of an adequate sensory phenomenology. 

What Sellars notoriously concluded was that sensory contents could be synoptically 
integrated into the scientific image only if both they and the micro-physical particulars 
of that image as well were subjected to yet another ontological reconception in terms 
of a categorially-monistic framework whose basic entities were all "absolute processes." 
Only when perceivers themselves had been reconceived as systems or "harmonies" of 
absolute processes, including the ultimate conceptual descendants of sensory contents, 
would the way be cleared for a unification of the two images. Thought of as absolute 
processes, sensings would be physical 

not only in the weak sense of not being mental (i.e. conceptual), for they lack intention- 
ality, but in the richer sense of playing a genuine causal role in the behavior of sentient 
organisms. They would ... be physical-1 but not physical-2. Not being epiphenomenal, 
they would conform to a basic metaphysical intuition: to be is to make a difference. 
(1981,111: 126) 

Intention and action 

In contrast to the integrative challenges posed by thoughts and sensations, the 
challenge of integrating actions properly so called, that is, conducts informed by inten- 
tion and volition, into the scientific image is not fundamentally ontological. Although 
they exhibit quite special features when considered functionally, regarded from the 
ontological perspective, intentions and volitions are simply species of occurrent 
thought-episodes. What makes such thoughts practical are their special relationships 
to conduct or behavior, analogous to the way in which their status as non-inferential 
responses to sensations confers on particular thoughts the functional role of percep- 
tual judgments. 

Sellars characteristically signals the special conduct-determining role of practical 
thoughts by a contrived use of the auxiliary verb "shall" as an operator on thought 
contents expressed as sentences. Categorical intentions are temporally determinate 
first-person future-tensed practical thoughts. They have the canonical form (INT): Shall 
(I will do X at t). Volitions ("acts of will") are special cases of such intentions, whose 
time determination is the indexical present, thus, (VOL): Shall (I will now do X). 

Such practical thinkings, on Sellars's view, mediate between deliberative reasoning 
and overt behavior by being appropriately caught up in a network of acquired causal 
propensities that guarantee, roughly, that intentions of the form (INT) regularly give 
rise, at time t, to volitions of the form (VOL), which, absent paralysis and the like, in 
turn regularly give rise, then and there, to bodily movements that are (further circum- 
stances being appropriate) the initial stages of a doing of X. Such practical thinkings 
are governed according to a single principle which unites practical and theoretical 
reasoning: If p implies q[ then ' Shall(p) ' implies Shall(q) . 




Here again, then, Sellars concluded, what the manifest image contains is not the concept 
of something with a determinate intrinsic character with which we are acquainted but 
rather the functional conception of a causally-mediating logico-semantic role-player. 
Practical thinkings can thus be ontologically accommodated within the scientific 
image along the lines already sketched for cognitive thoughts in general. But here, he 
continued, ontological accommodation cannot be the end of the synoptic story, ff we 
take seriously the idea that the scientific image purports to be a complete image of man- 
in-the-world and a candidate ultimately to replace the manifest image, then the latter's 
categories pertaining to persons will need to reappear within the sought synoptic fusion 
as such. We need to reconcile "the idea that man is what science says he is" with "the cat- 
egories pertaining to man as a person who finds himself confronted by standards (ethical, 
logical, etc.) which often conflict with his desires and impulses" (1963c: 38). 

On Sellars's view, the basic concept of a person is irredeemably social. To think of 
an entity as a person is essentially to think of it as actually or potentially a member of 
a community, "an embracing group each member of which thinks of itself as a member 
of the group" (1963c: 39), and it is the most general shared intentions of its members 
that fundamentally define the structure of norms and values in terms of which their 
conducts come to be appraised as "correct" or "incorrect" or "right" or "wrong." 

Thus the conceptual framework of persons is the framework in which we think 
of one another as sharing the community intentions which provide the ambience 
of principles and standards (above all, those which make meaningful discourse 
and rationality itself possible) within which we live our own individual lives. (1963c: 

As we have seen, Sellars interpreted the framework of thoughts as founded within 
the manifest image on a series of ontologically noncommittal functional analogies to 
which we can readily imagine an emerging scientific understanding progressively sup- 
plying structural (e.g. neurophysiological) form. In contrast, he argued that accom- 
modating the manifest image's sensory contents within a synoptic fusion would require 
the conceptual transposition of some of its ontologically basic entities into new cate- 
gorial forms. Unlike the frameworks of thoughts and sensations, however, Sellars con- 
tended that the conceptual framework of persons as such "is not something that needs 
to be reconciled with the scientific image, but rather something to be joined to it" 
(1963c: 40). To achieve a genuinely synoptic vision of man-in-the-world, he con- 
cluded, the scientific image needs to be enriched 

not with more [or different] ways of saying what is the case, but with the language of com- 
munity and individual intentions, so that by construing the actions we intend to do and 
the circumstances in which we intend to do them in scientific terms, we directly relate the 
world as conceived by scientific theory to our purposes, and make it our world and no 
longer an alien appendage to the world in which we do our living. (1963c: 40) 

Such an ultimate unification of the manifest and scientific images, the world of 
persons with the world of science, was the controlling vision of Sellars's philosophy. 



What makes him one of the towering figures of postwar analytic philosophy, however, 
is not just the grand scope of his enterprise, but the profound originality of his specific 
conclusions, the sophisticated dialectically argued and historically informed reasoning 
with which he supported them, and the exemplary thoroughness with which he 
painstakingly developed something very rare in the analytic tradition, a principled and 
consistent systematic philosophical view. 

Bibliography of works by Sellars 

1963a: "Empiricism and the Philosophy of Mind," in 1963d, pp. 127-96. (Formerly published 

in The Foundations of Science and the Concepts of Psychoanalysis, Minnesota Studies in the 

Philosophy of Science, vol. I, ed. H. Feigl and M. Scriven, Minneapolis: University of Minnesota 

Press, 1956.) 
1963b: "Phenomenalism," in 1963d, pp. 60-105. 
1963c: "Philosophy and the Scientific Image of Man," in 1963d, pp. 1-40. (Formerly published 

in Frontiers of Science and Philosophy, ed. Robert Colodny, Pittsburgh, PA: University of 

Pittsburgh Press, 1962.) 
1963d: Science, Perception and Reality, London: Routledge and Kegan Paul, and New York: 

Humanities Press. (Reissued by Ridgeview Publishing.) 
1963e: "The Language of Theories," in 1963d, pp. 106-26. (Formerly published in Current 

Issues in the Philosophy of Science, ed. H. Feigl and G. Maxwell, New York: Holt, Rhinehart and 

Winston, 1961.) 
1967a: "Abstract Entities," in 1967b, pp. 229-69. Reprinted in Review of Metaphysics 16 (1983). 
1967b: Philosophical Perspectives, Springfield, IL: Charles C. Thomas. (Reissued in 2 vols, 

Philosophical Perspectives: History of Philosophy and Philosophical Perspectives: Metaphysics and 

Epistemology , by Ridgeview Publishing.) 
1967c: "The Identity Approach to the Mind-Body Problem," in 1967b, pp. 3 70-88. (Formerly 

published in Review of Metaphysics 18 (1965).) 
1968: Science and Metaphysics: Variations on Kantian Themes, London: Routledge and Kegan Paul 

and New York: Humanities Press. (Reissued by Ridgeview Publishing.) 
1974: "Meaning as Functional Classification," Synthese 27, pp. 417-37. 
1975a: "Empiricism and Abstract Entities", in 1975b, pp. 245-86. (Formerly published in The 

Philosophy of Rudolph Carnap, ed. P. A. Schilpp, La Salle, IL: Open Court, 1963.) 
1975b: Essays in Philosophy and its History, Dordrecht: Reidel. 
1975c: "Induction as Vindication," in 1975b, pp. 367-416. (Formerly published in Philosophy 

of Science 31 (1964).) 
1975d: "The Structure of Knowledge," in Action, Knowledge, and Reality: Studies in Honor of 

Wilfrid Sellars, ed. H. N. Castaneda, Indianapolis: Bobbs-Merrill, pp. 295-347. 
1975e: "Language as Thought and Communication," in 1975b, pp. 93-117. (Formerly pub- 
lished in Philosophy and Phenomenological Research 29 (1969).) 
1979: Naturalism and Ontology, Reseda, CA: Ridgeview Publishing. 
1 9 80: Pure Pragmatics and Possible Worlds: The Early Essays of Wilfrid Sellars, ed. J. F. Sicha, Reseda, 

CA: Ridgeview Publishing. (Also contains a long introductory essay by Sicha and an extensive 

bibliography of Sellars's work through 19 79.) 
1981: "The Carus Lectures for 1977-78," The Monist 64/1. (Citations by lecture and numbered 

1989: The Metaphysics of Epistemology, Lectures by Wilfrid Sellars, ed. P. Amaral, Reseda, CA: 

Ridgeview Publishing. (Also contains a complete bibliography of Sellars's published work 

through 1989.) 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

H. P. Grice (1913-1988) 



Herbert Paul Grice was born on March 13, 1 9 1 3 , in Birmingham, England. He attended 
Corpus Christi College, Oxford, graduating in 1936. From 1938 until 1967 he held 
various fellowships and lectureships at St John's College. His time at Oxford was inter- 
rupted by nearly five years' wartime service in the Royal Navy, first in the North Atlantic 
and later in Admiralty intelligence, in 1967, he moved to the University of California, 
Berkeley as Professor of Philosophy. He was elected to the British Academy in 1966, 
and gave the William James Lectures at Harvard in 1967, the John Locke lectures at 
Oxford in 1978, and the Tanner Lectures at Stanford in 1980. He died in Berkeley in 
August 1988, shortly before the publication of his first book, Studies in the Way of 

Grice was one of the most gifted and respected philosophers of the second half of 
the twentieth century. He set impossibly high standards and was always reluctant to go 
into print - heroic efforts were required by editors and friends to extract from him the 
handful of papers he deemed worthy of publication - yet he exerted considerable influ- 
ence through seminars and invited lectures. He worked on topics in Aristotle, meta- 
physics, ethics, and philosophical psychology; but his strongest influence was in the 
philosophy of language, where his thought continues to shape the way philosophers, 
linguists, and cognitive scientists think about meaning, communication, and the rela- 
tion between language and mind. 

With respect to a particular sentence, X, and an "utterer" U, Grice stressed the 
importance of separating (1) what X means, (2) what U said on a given occasion by 
uttering X, and (3) what U meant by uttering Xon that occasion. Second, he attempted 
to say what meaning is by providing analyses of utterer's meaning, sentence meaning, 
and what is said. Third, he tried to explain how what U says and what U means can 
diverge. Fourth, he defended conceptual analysis and some form of analytic/synthetic 
distinction. Fifth, by characterizing the distinction between the "genuinely semantic" 
and "merely pragmatic" implications of a statement, Grice clarified the relationship 
between classical logic and the semantics of natural language. Sixth, he deployed his 
notion of "implicature" to devastating effect against overzealous strains of "ordinary- 
language philosophy," without abandoning the view that philosophy must pay atten- 



tion to the nuances of ordinary talk (see AUSTIN). Seventh, Grice undercut the most 
influential arguments for a philosophically significant notion of "presupposition." 
Eighth, he made significant contributions to debates about the semantics of proper 
names, definite descriptions, and pronouns. Ninth, he sketched a philosophical psy- 
chology and a theory of value that promise to provide the basis of future work on 
actions, mental states, and moral philosophy, and to explain the relationship between 
mind and language inherent in his philosophy of language. 

Meaning, use, and ordinary language 

The view that the only useful thing to say about the meaning of an expression is that 
it is usable in such-and-such circumstances, exercised a powerful influence on philoso- 
phy in postwar Oxford. Austin, Ryle, and others undercut philosophical positions or dis- 
posed of philosophical problems by pointing to a misuse of some expression playing an 
essential role in the presentation of the position or problem. Consider attempts to 
analyze knowledge in terms of belief along the following lines: A knows thatp if and only 
if (1) A believes that p, (2)p, and (3) A is justified in believing that p. It might be charged 
that it is a feature of the use of "believe" that one does not use it if one can sincerely 
use "know" instead. Such a claim might be supported by observing that it would be 
inappropriate for a man to say "I believe Smith is dead" when he knows Smith is dead. 
And so it might be concluded that the proposed analysis must be discarded because 
clause (1) conflicts with the ordinary use of the verb "believe." 

Grice accepted that a theory of meaning must be sensitive to use and attempted to 
explicate the meaning of an expression (or any other sign) in terms of what its users 
do with it, that is, in terms of what its users (could/would/should) mean by it on par- 
ticular occasions of use. Two important ideas came out of this sensitivity to use. The 
first is that the locution by uttering x, U meant that p can be analyzed in terms of complex 
audience-directed intentions on the part of U. The second is that the most "basic" 
notion of meaning is that of an utterer U meaning something by doing something on 
a particular occasion; all other notions of meaning are derivative. What U means by 
producing x on a given occasion is a function of what U intends, in a complex way, to 
get across to his audience. The basic idea is, very roughly, that for an "indicative-type" 
utterance, the locution by uttering x, U meant that p expresses a truth iff U uttered x 
intending to produce in some audience A the belief thatp by means of A's recognition 
of this intention. Sentence meaning is to be analyzed in terms of regularities over the 
intentions with which utterers produce sentences on given occasions. 

By uttering a sentence of the form "p or q," U may well imply that he has non-truth- 
functional grounds for his assertion; but this is not part of what the sentence (or the 
statement made) implies. Grice wanted any adequate explanation of the possibility of 
pragmatic implications to flow from a completely genera] theory. To demonstrate the def- 
inite existence of pragmatic implications distinct from semantic implications, Grice con- 
sidered an extreme example. Suppose A asks U for an evaluation of his student Mr. X. 
All U says is "Mr. X has excellent handwriting and is always very punctual." ff U leaves 
it at that, those present are likely to conclude that U thinks Mr. X is not much good at 
philosophy. There is surely no temptation to say that the proposition that Mr. X is not 
much good (or that U thinks Mr. X is not much good) at philosophy is (or is a conse- 



quence of) the statement U made. The sentence U uttered has a clear linguistic meaning 
based on the meanings of its parts and their syntactical arrangement; and it seems 
quite wrong to say that, when he uttered that sentence, U made the statement that Mr. 
Xis not much good at philosophy. On the other hand, it seems quite natural to say that, 
in the circumstances, what U meant (or part of what U meant) by making the state- 
ment he in fact made was that Mr. X is not much good (or that U thinks Mr. X is not 
much good) at philosophy. This is something the utterer implied by making the state- 
ment he did in this context, not something implied by the sentence uttered or by the 
statement U made by uttering the sentence. 

The theory of conversation 

With respect to what U means by a linguistic utterance, Grice proposed to separate what 
U says and what U implicates (e.g. implies, indicates, or suggests). What 17 says is to be 
closely tied to the conventional meaning of the words uttered, which both falls short 
and goes beyond what is said. 

It falls short because a specification of what U said on a particular occasion must 
take into account not only the conventional meaning of the sentence used but also (e.g.) 
the references of referring expressions (e.g. proper names, demonstratives, and indexi- 
cals) and the time and place of utterance. What U said is to do duty for what U stated or 
the proposition expressed by U. Where the sentence uttered is of the type conventionally 
associated with the speech act of asserting (i.e. when it is in the "indicative mood") 
what is said will be straightforwardly truth-conditional. When the sentence uttered is in 
the imperative or interrogative mood, what is said will not be straightforwardly truth- 
conditional, but it will be systematically related to the truth conditions of what U would 
have said, in the same context, by uttering the indicative counterpart (or one of the 
indicative counterparts) of the original sentence. 

The conventional meaning of a sentence also goes beyond what is said because of 
devices that signal the performance of "noncentral speech acts" parasitic upon the per- 
formance of the "central speech acts" of asserting, questioning, and ordering. Such 
devices, although they play a part in determining what U meant, play no part in deter- 
mining what U said. If U utters (1) rather than (2), 

1 She is poor but she is honest 

2 She is poor and she is honest 

very likely U will be taken to be implying that there is (or that someone might think 
there is) some sort of contrast between poverty and honesty (or her honesty and her 
poverty). This type of implication is no part of what U says because it does not con- 
tribute in any way to the truth conditions of the utterance. By uttering (1), U has said 
only that she is poor and she is honest; and this does not entail that there is any (e.g.) 
contrast between poverty and honesty (or between her poverty and her honesty). The 
implication in question Grice calls a conventional implicature. 

According to Grice, by uttering (1) 17 is performing two speech acts, saying that she 
is poor and she is honest and indicating (or suggesting) that someone (perhaps U) has a 
certain attitude toward what is said. Grice did not develop this idea; he just left us with 
the claim that a conventional implicature is determined (at least in part) by the (con- 



what U 

what U 



what U 



what U 

what U 

what U 

what 17 








Figure 1 

ventions governing) the words used. He does stress, however, that the sort of implica- 
tion we have just been considering is not a presupposition (as originally defined by 
Strawson and adopted by others). B is a presupposition of A, just in case the truth 
or falsity of A requires the truth of B. (If the truth of A requires the truth of B, but 
the falsity of A does not, B is an entailment of A.) More precisely, if A presupposes B, A 
lacks a truth value if B is false. But as Grice points out, an utterance of (1) can be 
false even if the implied proposition is false, effectively scotching the idea that the impli- 
cation is presupposition (at least not on the standard semantic conception of that 
notion), ft is Grice's view that any alleged presupposition is either an entailment or 
an implicature. 

For something to be (part of) what U says, it must also be (part of) what U meant, 
that is, it must be backed by a complex intention of the sort that forms the backbone 
of Grice's theory of meaning (see figure 1). ff U utters the sentence "Bill is honest" iron- 
ically, on Grice's account U will not have said that Bill is honest: U will have made as if 
to say that Bill is honest. For it is Grice's view that a statement of the form "by uttering 
x, U said that p" entails the corresponding statement of the form "by uttering x, U 
meant that p." So on Grice's account, one cannot unintentionally say something (a fact 
that has interesting consequences for, for example, slips of the tongue and misused 

Grice's work provides a breakdown of what U meant as shown in figure 1. 

What U conventionally implicates and what U says are both closely tied to the con- 
ventional meaning of the sentence uttered, and they are taken by Grice as exhausting 
what U conventionally means (i.e. means by virtue of linguistic convention). Let us now 



turn to what 17 non-conventionally means. Consider again, the example concerning 
Professor 17' s evaluation of Mr. X. By uttering the sentence "Mr. Xhas excellent hand- 
writing and is always very punctual," U said (or made as if to say) that Mr. Xhas excel- 
lent handwriting and is always very punctual. In addition, on Grice's account 77 
conversationally implicated that Mr. Xis not much good at philosophy (there is a conver- 
sational implicature to the effect that Mr. X is not much good at philosophy). 
Conversational implicature is a species of pragmatic (non-semantic, non-conventional) 
implication and is to be contrasted with the (at least partly semantic) implication that 
Grice calls conventional implicature. The principal difference between a conventional 
and a conversational implicature is that the existence of a conventional implicature 
depends upon the presence of some particular conventional device (such as "but," 
"moreover," "still," "yet," or heavy stress) whereas the existence of a conversational 
implicature does not. 

Grice proposes to explain the possibility of a divergence between what 17 says and 
what 17 means (or at least between what 17 conventionally means and what 77 means) 
by appeal to the nature and purpose of rational interaction. Conversation is viewed by 
him as a characteristically purposeful and cooperative enterprise governed by what he 

The Cooperative Principle: Make your conversational contribution such as is required, 
at the stage at which it occurs, by the accepted purpose or direction of the talk exchange 
in which you are engaged. Subsumed under this principle, Grice distinguishes four 
categories of more specific maxims and submaxims enjoining truthfulness, informa- 
tiveness, relevance, and clarity: 

Quantity: Make your contribution as informative as is required (for the current 
purposes of the exchange): do not make your contribution more informative than is 

Quality: Try to make your contribution one that is true. Specifically: (1) Do not say 
what you believe to be false; (2) Do not say that for which you lack adequate evidence. 

Relation: Be relevant. 

Manner: Be perspicuous. Specifically: (1) Be brief; (2) Be orderly; (3) Avoid ambigu- 
ity; (4) Avoid obscurity of expression. 

Grice's basic idea is that there is a systematic correspondence between what 17 
means and the assumptions required in order to preserve the supposition that U is 
observing the Cooperative Principle and conversational maxims. In the case of 
Professor L7's evaluation of Mr. X, on the surface the Cooperative Principle or one 
or more of the maxims is intentionally and overtly not fulfilled. By saying "Mr. X has 
excellent handwriting and is always very punctual" (in this particular context), 77 
seems not to have fulfilled one of the maxims of Quantity or the maxim of Relation. 
(If Mr. X is one of U's students, 77 must be in a position to volunteer more relevant 
information than judgments about Mr. Xs handwriting and timekeeping; furthermore, 
U knows that more information, or more relevant information, is required.) The 
hearer is naturally led to the conclusion that 77 is trying to convey something 
else, something more relevant to the purposes at hand. In the circumstances, if 
U thought Mr. X was any good at philosophy he would have said so. So 77 must 
think Mr. X is no good at philosophy and be unwilling to say so. And so 77 has conver- 
sationally implicated that Mr. X is no good at philosophy. 



One interesting feature of this example is that it might well be the case that 
only what is implicated is meant (i.e. backed by U's communicative intentions). U 
may have no idea what Mr. X's handwriting is like because Mr. X has shown U only 
typed manuscripts of his work (or because he has never shown U anything), and U 
may have no opinion as to whether or not Mr. X is punctual. In such a version of the 
envisioned scenario, U has only made as if to say that Mr. X has excellent handwriting 
and is always very punctual because U had no intention of inducing (or activating) 
in his audience the belief that (U thinks that) Mr. X has excellent handwriting and 
is always very punctual. The truth-values of what U said (or made as if to say) 
and what U conversationally implicated may of course differ. Mr. X may have quite 
atrocious handwriting, and U may know this; but given the relevance of what is 
conversationally implicated, U may care very little about the truth-value of what 
he has said (or made as if to say). The primary message is to be found at the level of 
what is conversationally implicated. 

In general, Grice claims, a speaker conversationally implicates that which he must 
be assumed to think in order to maintain the assumption that he is observing the 
Cooperative Principle (and perhaps some conversational maxims), if not at the level 
of what is said, at least at the level of what is implicated. At some overarching level of 
what is meant, U is presumed to be observing the Cooperative Principle. The wording of 
Grice's maxims suggests that some concern only what is said (e.g. "Do not say what you 
believe to be false") while others concern, perhaps, what is meant (e.g. "Be relevant"). 
We should probably treat this as something of an uncharacteristic looseness of expres- 
sion on Grice's part. Except for the maxims under Manner (which can apply only to 
what is said) it seems reasonable to understand Grice as allowing a maxim not to be 
fulfilled at the level of what is said to be licensed or overridden by adherence at the level 
of what is implicated. On such a view, blatantly violating a maxim at the level of what 
is said but adhering to it at the level of what is implicated would not necessarily involve 
a violation of the Cooperative Principle. 

Some important questions are still unanswered: How are saying and implicating 
to be defined? How are implicatures calculated? What is the status of the Cooperative 
Principle and maxims? What happens when a speaker cannot simultaneously observe 
all of the maxims? It is important to see how Grice attempts to face such questions. 

No one should deny that in the example of the evaluation of Mr. X there is an 
intuitive and obvious distinction to be made between what U said and what U conver- 
sationally implicated. But in view of the sorts of example that really bother Grice - 
"the F is G," "p or q," "if p then q," etc. - he could not rest with an intuitive distinction. 
The example concerning the evaluation of Mr. X is clear-cut, obvious, and uncon- 
tentious. And herein lies the problem. The examples of purported conversational 
implicature that most interest Grice are philosophically important ones with respect 
to which many philosophers have not felt the need to invoke such a distinction. This 
might be because it is not at all obvious that there is such a distinction to be 
made in the cases in question (or if there is, how relevant it is), or because adherence 
to some form of the "meaning is use" dogma has blinded certain philosophers to 
the possibility of such a distinction. So Grice ultimately needs analyses of "what is said" 
and "what is conversationally implicated" in order to get philosophical work out 
of these notions. 



Grice hopes to analyze the notion of saying in terms of utterers' intentions. This pro- 
posal will be examined after discussion of Grice's theory of meaning. He attempts to 
define conversational implicature in terms of an, as yet, undefined notion of saying. 
The following schema is supposed to be a first step: 

Someone who, by (in, when) saying (or making as if to say) that p has implicated 
that q, has conversationally implicated that q, provided that (f) he is to be presumed to 
be observing the conversational maxims, or at least the Cooperative Principle; (2) the 
supposition that he is aware that q is required in order to make his saying or making as 
if to say p consistent with this presumption; and (3) the speaker thinks (and would 
expect the hearer to think that the speaker thinks) that it is within the competence of 
the hearer to work out, or grasp intuitively, that the supposition mentioned in (2) is 

We appear to have here a set of necessary conditions. The conditions are not 
sufficient because conventional implicatures are not excluded. Whenever there is a 
conversational implicature, one should be able to reason somewhat as follows: (i) U 
has said thatp; (ii) there is no reason to suppose that U is not observing the Cooperative 
Principle and maxims; (iii) U could not be doing this unless he thought that q; (iv) U 
knows (and knows that I know that U knows) that f can see that U thinks the suppo- 
sition that U thinks that q is required; (v) U has done nothing to stop me thinking 
that q\ (vi) U intends me to think, or is at least willing to allow me to think, that q; (vii) 
and so, U has implicated that q. In each of the cases Grice considers, it does seem to be 
possible to justify the existence of the implicature in question in this sort of way. But 
notice that q is simply introduced without explanation in step (iii), so Grice has certainly 
not stated any sort of method or procedure for calculating the content of conversa- 
tional implicatures. A good deal of work needs to be done on the calculation of 
particular implicatures if Grice's evident insights are to form the basis of a finally 
acceptable theory. 

A necessary condition on conversational implicatures that is intimately connected 
to condition (3) is that they are intended. This follows, if not from condition (3), at least 
from the fact that (a) what U implicates is part of what U means, and (b) what U means 
is determined by U's communicative intentions. A hearer may think that, by saying that 
p, U has conversationally implicated that q (A may even have reasoned explicitly in the 
manner of (i)-(vii) above). But if U did not intend the implication in question it will not 
count as a conversational implicature. 

We have, then, four conditions that are necessary but not sufficient for classifying 
an implication as a conversational implicature. Entailments do not seem to have been 
excluded. In order that we may stay focused on the relation between the speaker and 
certain propositions, let us make a harmless addition to Grice's terminology. If the 
proposition thatp entails the proposition that q, then if U is a competent speaker who 
says thatp, U thereby says* that q. So if U is a perfectly competent English speaker who 
has sincerely uttered the sentence "John is a bachelor," not only has 17 said (and said*) 
that John is a bachelor, he has also said* that John is unmarried. It seems desirable that 
no proposition be both an entailment and a conversational implicature of the same 
utterance. But it is not obvious that the conditions laid down thus far on conversational 
implicature actually rule out entailments. Furthermore, Grice cannot just impose a 
further condition to the definition to the effect that no entailment is a conversational 



implicature. One of Grice's avowed aims is to ward off certain ordinary language argu- 
ments by invoking a sharp distinction between what we are now calling conversational 
implicature and entailment; so it is not good enough for him to use the notion of an 
entailment in a definition of conversational implicature. 

A fifth condition Grice imposes on conversational implicatures seems to help. Unlike 
an entailment, a conversational implicature is supposed to be cancelable either explic- 
itly or contextually, without contradiction, ff U says* that p, and p entails q, then U 
cannot go on to say* that not-q without contradiction. For example, U cannot say "John 
is a bachelor and John is married." But if U says* thatp, and thereby conversationally 
implicates that q, U can go on to say* that not-q without contradiction. Consider again 
the case of U's evaluation of Mr. X. After uttering "Mr. Xhas excellent handwriting and 
is always very punctual," U might (without irony) continue "Moreover, Mr. X's recent 
modal proof of the immortality of the soul is a brilliant and original contribution to 
philosophy." fn the light of the first comment, this addition might be rather odd, but it 
would not result in U contradicting himself. (In addition to distinguishing conversa- 
tional implicatures from entailments, the cancelability test is also supposed to distin- 
guish conversational from conventional implicatures. Although it will not lead to 
contradiction, attempting to cancel a conventional implicature will result in a gen- 
uinely linguistic transgression of some sort. This is precisely because there is a distinct 
semantic component to conventional implicatures.) 

Putting these five conditions together, we come as close as we can with Grice's 
machinery to a set of necessary and sufficient conditions on conversational 

Philosophical psychology 

For Grice, the principles involved in an account of conversational implicature are to be 
grounded in a philosophical psychology that explicates the purportedly hierarchical 
relationships that hold between the various types of psychological states we ascribe to 
creatures that can reason and form complex intentions. The beginnings of this line of 
thought can be traced to the end of his 1957 paper "Meaning." It contains the seeds 
of (1) the view that the Cooperative Principle and conversational maxims (in particu- 
lar the maxim enjoining relevance) are to play a central role not only in an account of 
possible divergences between what U said and what U meant but also in an account of 
the resolution of ambiguities, and (2) the view that the use of language is one form 
of rational activity and that the principles at work in the interpretation of linguistic 
behavior are (or are intimately related to) those at work in interpreting intentional 
non-linguistic behavior. 

Two questions spring to mind immediately: (1) What are the relative rankings of the 
maxims in cases where it is hard (or impossible) for U to observe all of them (or all of 
them to the same degree), and why? (2) What is the basis for the assumption that speak- 
ers will in general (ceteris paribus and in the absence of indications to the contrary) 
proceed in the manner prescribed by the Cooperative Principle and maxims? 

Grice is explicit about the position of at least one of the maxims of Quality in any 
hierarchy. Suppose A is planning an itinerary for a vacation to France. A wants to see 
his friend C, if so doing would not require too much additional traveling. A asks B 



"Where does C live?" B replies "Somewhere in the south of France." B knows that A 
would like more specific information but he is not in a position to be more specific. So 
B is faced with not fulfilling either a maxim of Quality or a maxim of Quantity. Quality 
wins out. The maxims of Quality have a very special status within Grice's overall theory 
and Grice entertains the idea that the first maxim of Quality should be part of some 
broader background; the other maxims come into operation only on the assumption 
that the maxim of Quality is satisfied. The maxims of Quality (or at least the first maxim 
of Quality) should not be thought of as admitting of degree or varying across cultures. 
In some sense this is an empirical matter; but unlike the maxims of Quantity and 
Manner, it does not seem very plausible to suppose that there are thriving cultures in 
which standardly people do not behave (for particular reasons to be determined by 
anthropologists) as if they are observing the maxims of Quality. 

Grice was not satisfied with the idea that it is just a well-recognized empirical 
fact that people do behave in accordance with the maxims and the Cooperative 
Principle, that in childhood they learned to do so and have not lost the habit. He wanted 
to find a basis that underlies our behavior and believed it would have a moral dimen- 
sion: not only do we in fact behave in the required way, but it is reasonable for us to do 
so, and the practice is something we should not abandon given our common purposes 
or goals. Conversation is one among a range of forms of rational activity for Grice. 
Observance of the CP and maxims is reasonable (rational): anyone concerned about 
the goals central to communication must be expected to have an interest, given 
suitable circumstances, in participation in informational exchanges that will be 
profitable only on the assumption that they are conducted in general accordance 
with the CP and maxims. 

On Grice's view, value predicates such as "proper," "correct," "optimal," and "rele- 
vant" cannot be kept out of an account of rational activity because a rational creature 
is essentially a creature that evaluates. Whether a value-oriented approach to the inter- 
pretation of intentional behavior can be developed in a fruitful way remains to be seen. 
But as Grice's unpublished work on ethics and philosophical psychology becomes more 
widely available, there will likely be a resurgence of interest in the matter of the precise 
location of the theory of conversation within a larger scheme. 

The logic of natural language 

One task of semantics is to provide a systematic characterization of judgments con- 
cerning truth, falsity, entailment, contradiction, and so on. fn the light of theoretical 
considerations, an initial judgment of, say, entailment might be rejected on the grounds 
that the perceived implication is an implicature rather than an entailment. So far, we 
have considered only examples of what Grice calls "particularized" conversational 
implicature, examples in which there is no temptation to say that the relevant impli- 
cation is an entailment (or a "presupposition"). Of more philosophical interest 
are "generalized" conversational implicatures, the presence and general form of 
which depend little upon the particular contextual details. Examples discussed by Grice 
include those attaching to utterances of sentences containing intentional expressions 
like "look," "feel," and "try," and "logical" expressions such as "and", "or," "if," "every," 
"a," and "the." 



According to Grice, philosophers who see divergences in meaning between "formal 
devices" such as "&,"v," "z>," "(Vx)," "(3x)," and "(ix)" and their natural language 
counterparts tend to belong to one of two camps, which he calls "formalist" and "infor- 
malist." The informalist position is essentially the one taken by Strawson (and others 
of the "ordinary-language movement"). The formalist camp is dominated by positivists 
and others who view natural language as inadequate to the needs of the science and 
philosophy of an age of precision. A typical formalist recommends the construction of 
an "ideal" or "logically perfect" language such as the language of first-order quantifi- 
cation theory with identity (or some suitable extension thereof). Since the meanings of 
the logical particles are perfectly clear, using an ideal language, philosophers can state 
propositions clearly, clarify the contents of philosophical claims, draw the limits of 
intelligible philosophical discourse, draw the deductive consequences of sets of state- 
ments, and generally determine how well various propositions sit with each other. 

Grice views the formalists and informalists as mistaken in the assumption of seman- 
tic divergence. Both sides have taken mere pragmatic implications to be parts of the 
meanings of sentences of natural language containing "logical" expressions. The case 
of "and" highlights some important methodological considerations. Although it is 
plausible to suppose that "and" (when it is used to conjoin sentences) functions seman- 
tically just like "&," there are certainly sentences in which it appears to function rather 

1 Jack and Jill got married and Jill gave birth to twins. 

2 Nero yelled and the prisoner began to tremble. 

Someone who uttered (1) would typically be taken to imply that Jack and Jill 
got married before Jill gave birth to twins. And someone who uttered (2) would typically 
be taken to imply that Nero's yelling contributed in some way to the prisoner's trem- 
bling. Thus one might be led to the view that "and" is not always understood as "&," 
that it is (at least) three ways ambiguous between truth-functional, temporal, and 
causal readings. 

The postulation of semantically distinct readings looks extravagant and Grice sug- 
gests it is good methodological practice to subscribe to "modified Occam's razor": senses 
are not to be multiplied beyond necessity. Given the viability of the distinction between 
what is said and what is meant, if a pragmatic explanation is available of why a par- 
ticular expression appears to diverge in meaning in different linguistic environments 
(or in different conversational settings) then ceteris paribus the pragmatic explanation 
is preferable to the postulation of a semantic ambiguity. Grice's idea is that the impli- 
cation of temporal sequence attaching to an utterance of (1) can be explained in terms 
of the fact that each of the conjuncts describes an event (rather than a state) and the 
presumption that U is observing the submaxim of Manner enjoining orderly deliveries. 
It seems to be Grice's view, then, that by uttering (1) 17 will conversationally implicate 
(rather than say) that Jack and Jill got married before Jill gave birth to twins (if this is 
correct then what is conversationally implicated would appear to entail what is said in 
this case). Similarly, the implication of causal connection attaching to an utterance of 
(2) is apparently to be explained in terms of the presumption that the speaker is being 
relevant. Before looking at problems for this proposal, I want first to get clear about its 



Conversational explanations trump semantic ambiguities on grounds of theoretical 
economy and generality. A conversational explanation is free: the mechanisms 
appealed to are already in place and independently motivated. The generality lost 
by positing several readings of "and" is quite considerable. First, implications of (e.g.) 
temporal priority and causal connection attach to uses of the counterparts of "and" 
across unrelated languages. Second, implications of the same sorts would surely arise 
even for speakers of a language containing an explicitly truth-functional connective 
"&." Third, the same implications that attach to utterances of "p and q" would attach 
to an utterance of the two sentence sequence "p. q" not containing an explicit 
device of conjunction. On methodological grounds, then, pragmatic accounts of the tem- 
poral and causal implications in (1) and (2) are preferable to accounts that appeal 
to semantic ambiguity. 

Grice opposes postulating idiosyncratic pragmatic rules with which to derive gener- 
alized implicatures. Conversational implicatures must be explicable in terms of the 
Cooperative Principle and maxims, construed as general antecedent assumptions about 
the rational nature of conversation. To call an implicature "generalized" rather than 
"particularized" is only to acknowledge the fact that the presence of the implicature is 
relatively independent of the details of the particular conversational context, a fact that 
is to be explained by the cooperative nature of conversation. 

A second challenge to classical logic semantics came from Strawson, who challenged 
Russell on the grounds that the theory does not do justice to ordinary usage: speakers 
use descriptions to refer, not to quantify, and hence Russell's theory is open to a number 
of objections (see STRAWSON). But according to Grice, a number of Strawson's objec- 
tions can be defused by distinguishing sentence meaning, what is said, and what is 

In Grice's terminology, one of Strawson's main complaints against Russell is that his 
theory conflates the meaning of a sentence "the F is G" and what U says by uttering 
this sentence (and similarly the subsentential counterparts of these notions) and so 
cannot explain the fact that U may say different things on different occasions by utter- 
ing the same sentence. Grice is right that Strawson can get no mileage out of Russell's 
failure to separate sentence meaning and what is said in his discussions. Upon reflec- 
tion it is clear that Russell's concern is with what is said rather than sentence meaning. 
If Russell were being more precise, he would not say that the sentence "the F is G" is 
equivalent to the sentence "there is exactly one F and every F is G"; rather, he would say 
that what U says by uttering "the F is G" on a particular occasion is that there is exactly 
one F and every F is G (occurrences of "F" in the foregoing may, of course, be ellipti- 
cal). The fact that a description (or any other quantified noun phrase) may contain an 
indexical component ("the present king of France," "every man here," etc.) does not 
present a problem: all this means is that there are some descriptions that are subject to 
the Theory of Descriptions (see RUSSELL) and a theory of indexicality. Grice is surely right, 
then, that although we need a sharp distinction between sentence meaning and what 
is said (and their subsentential counterparts), Strawson's appeal to this distinction 
when challenging Russell is empty. 

Grice neatly disposes of the view that descriptions are ambiguous between Russellian 
and referential (or identificatory) readings. When a description is used to identify some- 
thing, what U means diverges from what U says. What U says is given by the Russellian 



expansion but U also intends to communicate information about some particular indi- 
vidual, and although this is part of what U means, it is not part of what U says. This 
provides a perfectly satisfactory account of what is going on when U uses a description 
that does not fit its target, but such cases are not needed to see Grice's distinctions at 
work. According to Grice, when a description is used in an identificatory way, there will 
always be a mismatch between what U says and what U means (even where the descrip- 
tion uniquely fits the individual the speaker intends to communicate information about) 
because what is said is, on Russell's account, analyzable as a quantificational pro- 
position, whereas what is meant will always include a singular or object-dependent 

Again, methodological considerations strongly favor the Gricean account of refer- 
ential usage over an account that posits a semantic ambiguity: (1) If we were taught 
explicitly Russellian truth conditions, referential usage would still occur; (2) exactly 
parallel phenomena occur with indefinite descriptions and other quantified noun 
phrases; (3) modified Occam's razor enjoins us to opt for the simpler of two theories, 
other things being equal. Subsequently, far more detailed defenses of Russell along 
Gricean lines have been proposed by other philosophers, but the debts these works owe 
to Grice are considerable. More generally, a debt is owed to Grice for rejuvenating the 
position that classical logic is a remarkably useful tool as far as the semantics of natural 
language is concerned. 

The theory of meaning 

Grice attempted to analyse or explicate what is said and what is implicated in terms of 
intention, belief, desire, and recognition. Analyzing locutions of the forms "X did Y 
intentionally," "X caused Y," "X is true," "X entails Y," and so on, has been seen by 
many philosophers as a central task of philosophy. Grice's analyses of "by uttering X, 
U meant that p," "X means 'p'," and "by uttering X, U said that p" seem to have a reduc- 
tive and explicative flavor in that it appears to be his view that locutions of the forms 
can be wholly explicated without appealing to semantical concepts. 

He begins with what people mean rather than with what this or that expression, sign, 
or action means, seeking to analyze this in terms of complex audience-directed inten- 
tions on the part of the utterer, and to analyze utterance-type meaning (e.g. sentence 
meaning and word meaning) in terms of utterer's meaning. 

Although Grice aims to neutralize many ordinary language maneuvers with his 
saying/implicating distinction, one of the driving forces behind his work is still the idea 
that the meaning of an expression is a function of what its users do with it. Abstracting 
away from certain details that I will get to later, the direction of analysis for Grice is 
shown in figure 2. 

The idea, then, is to begin by providing an analysis of (1) utterer's meaning, and 
then to use this analysis in an analysis of (2) utterance-type meaning. (3) What is said 
is then to be defined in terms of a near coincidence of utterer's meaning and utterance- 
type meaning (for certain utterance-types); and finally (4) conversational implicature 
is to be defined in terms of saying and utterer's meaning. 

Although Grice does not address this point directly, it is clear that the task of expli- 
cating the locution "by uttering x, U said that p" takes on some urgency for him, 



(2) utterance-type meaning 

(1) utterer's meaning — > (3) what is said 



(4) what is conversationally implicated 

The " — > " is understood as " (or its analysis) plays a role in the analysis of 

(but not vice versa)." 

Figure 2 

because the saying/implicating distinction is so central to his attempts to counter ordi- 
nary language arguments of the sort examined earlier. A direct analysis of saying 
appears out of the question because Grice openly declares that he is using "say" in a 
special sense, and this precludes systematic appeal to intuitions about ordinary usage. 
By contrast, when it comes to pronouncing on the truth of instances of "by uttering x, 
U meant thatp," Grice believes he can help himself to such intuitions, many of them 
quite subtle. Strictly speaking, then, saying is to be defined rather than analyzed. 

To some philosophers and linguists, Grice's program seems to constitute something 
of a snub to serious compositional semantics. The idea that sentence meaning is to be 
analyzed in terms of utterer's meaning has been felt to conflict with (1) the fact that 
knowing the meaning of a sentence is typically a necessary step in working out what 
U meant by uttering that sentence, i.e. for recovering ITs communicative intentions, 
and (2) the fact that the meaning of a sentence is determined, at least in part, by the 
meanings of its parts (i.e. words and phrases) and the way the parts are put together 
(syntax). Both of these charges are based on misunderstandings of Grice's project, as 
will become clear. 

Utterer's meaning 

The basic Gricean analysis of utterer's meaning is this: 

I. "By uttering x, U meant something" is true iff for some audience A, U uttered x 

(1) A to produce some particular response r, 

(2) A to recognize that 17 intends (1), and 

(3) A's recognition that U intends (1) to function, in part, as a reason for (1). 

To provide a specification of r, says Grice, is to say what U meant. Where x is an 
"indicative" utterance, r is As believing something. 
If. "By uttering x, U meant that p" is true iff for some audience A, U uttered x 

( 1 ) A to believe that p, 

(2) and (3) as above. 



This type of complex intention Grice calls an "M-intention": by uttering x, U meant 
that p iff for some audience A, U uttered x M-intending A to believe that p. 

Two general problems face II. The first is that Grice provides a number of examples 
in which it would be correct to say that U means that p but incorrect to say that U 
intends A to believe that p (1989: 105-9). Suppose U is answering an examination 
question and says "The Battle of Waterloo was fought in 1815." Here U meant that the 
Battle of Waterloo was fought in 1815; but U did not M-intend the examiner to think 
that the Battle of Waterloo was fought in 1815 (typically, U will be under the impres- 
sion that the examiner already knows the answer). In response to this and related 
examples, Grice suggests that clause (1) of II. be changed to (1 : ): 

(lx) A to think that 17 thinks that p. 

A distinction is then made between exhibitive utterances (utterances by which U M- 
intends to impart the belief that U has a certain propositional attitude) and protreptic 
utterances (utterances by which U M-intends, via imparting a belief that [U] has a 
certain propositional attitude, to induce a corresponding attitude in the hearer). 

The suggested revision may not seem to comport with the commonly held view that 
the primary purpose of communication is the transfer of information about the world; 
on the revised account, the primary purpose seems to be the transfer of information 
about one's mental states. Another worry is that even if the proposed revision is an 
improvement, it does not weaken the analysis in such a way as to let in cases of remind- 
ing (some cases of which bring up another problem). Suppose U knows that A thinks 
thatp but needs reminding. So U does something by which he means that p. Not only 
does it seem incorrect to say (as the original analysis would require) that U intends A 
to think that p - U knows that A already thinks that p - it also seems incorrect to say 
(as the modified analysis requires) that U intends A to think that U thinks that p ([/may 
know that A already thinks that U thinks that p). What seems to be needed here, says 
Grice, is some notion of an activated belief: (1) needs to be changed not to (lj) but to 
something more like (1 2 ): 

(1 2 ) A actively to believe that U thinks that p. 

But there seems still to be a problem involving reminding. Suppose A has invited B 
over for dinner tonight at seven-thirty. B has agreed to come but U doubts B will show 
up and says as much to A. At seven o'clock, U and A are deep in philosophical conver- 
sation and U, realizing that A has lost track of time, says "B will be here in half an 
hour." This type of example suggests we are better off with something like (1 3 ), at least 
for some cases: 

(1 3 ) A actively to believe that p. 

So perhaps a disjunctive clause is going to be required in any finally acceptable analysis. 
Perhaps the problem with the first clause of II. is an instance of a more general dif- 
ficulty concerning the content of the intention (or M-intention) characteristic of com- 
municative behavior. This seems to be the view of Searle (1969). One way of putting 
Searle's general point is as follows: by paying too much attention to examples in which 
U intends to induce in A some propositional attitude or other, Grice has mistakenly 
taken a particular type of intention that does in fact accompany many utterances - the 



subintention specified in clause (1) - to be an essential ingredient of communicative 
behavior. But there are just too many cases of meaning involving linguistic (or other- 
wise conventional) utterances in which U does not seek to induce in an audience any 
propositional (or affective) attitude. Searle brings up three problems: first, it is not at all 
clear what attitude f M-intend to impart when making a promise by uttering a sentence 

of the form "I promise to "; second, sometimes I don't care whether I am believed or 

not; I just feel it is my duty to speak up; third, only an egocentric author intends me to 
believe that p because he has said so. 

These are genuine difficulties for Grice's analysis as it stands, but they do not seem 
to warrant abandoning Grice's project; rather they suggest that the specification of the 
type of response mentioned in the first clause needs to be weakened to something like 
the following: 

(1 4 ) A actively to entertain the belief/thought/proposition that p. 

Of course, in many cases U also intends (or at least would like) A to go on to believe 
that p, but this fact would not enter into the analysis of utterer's meaning. A revision 
along these lines might provide the beginning of a way out of Searle's problems. 

The second problem is that clause (3) of II. seems problematic. The original motiva- 
tion for clause (2) is clear. It is not enough, Grice points out, for U to mean thatp, that 
U utter x intending A to think that p. U might leave B's handkerchief near the scene of 
the murder with the intention of getting the detective (actively) to entertain the 
thought that B is the murderer. But there is no temptation to say that by leaving the 
handkerchief, U meant that B is the murderer. Hence clause (2), which requires U to 
intend A to recognize the intention specified in the first clause (however stated). 

But what of clause (3)? Grice wants this in order to filter out cases in which some 
natural feature of the utterance makes it completely obvious that p. He worried about 
cases like this: in response to an invitation to play squash, Bill displays his bandaged 
leg. According to Grice, we do not want to say that Bill meant that his leg was bandaged 
(though we might want to say that he meant that he could not play squash, or even 
that he had a bad leg). 

Many people's intuitions are less robust than Grice's here. He seems to be worried 
that in cases like these there is something approximating natural meaning that inter- 
feres with the idea of Bill non-naturally meaning that he has a bandaged leg. Given the 
links Grice seeks to forge between natural and non-natural meaning, it is not clear why 
the putative presence of natural meaning is supposed to be problematic, and so it is not 
clear why the third clause of II. is needed. Grice himself brings up cases that seem to 
create a problem for the third clause. Suppose the answer to a certain question is "on 
the tip of A's tongue." U knows this; that is, U knows that A thinks that p but can't 
quite remember. So [/reminds A thatp by doing something by which he (U) means that 
p. In such a scenario, even if U has the intention specified in the first clause (however 
stated), it does not seem to be the case that U has the intention specified in the third 
clause. It is noteworthy that the examples Grice uses to justify the third clause involve 
non-linguistic utterances (Grice's "John the Baptist" and "bandaged leg" cases). 
However, it is possible to construct cases involving properly linguistic utterances in 
which the fact thatp is made just as obvious by the utterance as in Grice's non-linguistic 
cases. Consider an utterance by me of (e.g.) "I'm right here" yelled in the direction of 



someone known to be looking for me. Here there is a strong inclination to say that I did 
not mean what I said. 

Problems await Grice if he does not concede the third clause is overly restrictive. 
Ultimately, he wants to define locutions of the form "by uttering x, U said thatp"; but 
one of the conjuncts in his proposed definiens is "by uttering x, U meant that p." So if 
he refuses to allow that (e.g.) I can mean that I can speak in a squeaky voice by utter- 
ing, in a squeaky voice, "I can speak in a squeaky voice," Grice will be forced either to 
conclude that I have not said that f can speak in a squeaky voice, or else to abandon the 
idea of defining saying in terms of utterer's meaning (he cannot, of course, say that in 
such a scenario I have only "made as if to say" that I can speak in a squeaky voice). It 
would seem, then, that the third clause will have to be discarded (or at least modified) 
if saying requires meaning. 

One positive result of discarding the third clause would be the disappearance of 
the "tip-of-the-tongue" problem. Another would be that Bill could mean that he had 
a bandaged leg in the scenario above, which is not obviously incorrect. When it 
comes to linguistic utterances, there might well be another interesting consequence. 
Typically, linguistic utterances do not seem to be underwritten by intentions as complex 
as M-intentions. Weakening the analysans by the removing clause (3) goes a long 
way toward quieting this worry; however, there are grounds for thinking that the 
relevant intention will have to be more complex than the one specified by clauses (1) 
and (2). 

The following type of example shows that clauses (1), (2), and (3) do not specify 
a rich enough intention (or batch of intentions). Suppose A, a friend of mine, is 
about to buy a house. I think the house is rat-infested, but I don't want to mention 
this outright to A so I let rats loose in the house knowing that A is watching me. I 
know that A does not know that I know that he is watching me do this. I know A 
will not take the presence of my rats to be natural evidence that the house is 
rat-infested; but I do know, indeed I intend, that A will take my letting rats loose in 
the house as grounds for thinking that I intend to induce in him the belief that 
the house is rat-infested. Conditions (l)-(3) of II. above are fulfilled. But surely it 
is not correct to say that by letting rats loose in the house I mean that the house is 

The problem is that in this example my intentions are not wholly overt. One possible 
remedy involves adding a fourth clause: 

(4) A to recognize that U intends (2). 

But the same sort of counterexample can still be generated, and then we need a fifth 
clause, then a sixth, and so on. Grice proposed to block an infinite regress by adding a 
condition that would prohibit any "sneaky" intention: instead of adding additional 
clauses, his idea was to add a second part to the analysis, the rough import of which is 
that U does not intend A to be deceived about U's intentions (l)-(3). As long as U does 
not have a deceptive intention of this sort, U is deemed to mean that p. 

Something like the following is best seen as the characterization of utterer's meaning 
that Grice left us to explore and refine: 

III. By uttering x, U meant thatp iff for some audience A, 



( 1 ) U uttered x intending A actively to entertain the thought that p (or the thought 
that U believes that p) 

(2) U uttered x intending A to recognize that U intends A actively to entertain the 
thought thatp 

(3) U does not intend A to be deceived about U's intentions (1) and (2). 

Sentence meaning and saying 

The idea of using utterer's meaning to explicate sentence meaning is thought by some 
philosophers to conflict with the idea that the meaning of a sentence is a function of 
the meanings of its parts (i.e. words and phrases) and their syntactical organization. 
Grice's project gets something "backwards" it is claimed: surely any attempt to model 
how we work out what someone means on a given occasion will progress from word 
meaning plus syntax to sentence meaning, and from sentence meaning plus context to 
what is said, and from what is said plus context to what is meant. And this clashes with 
Grice's view that sentence meaning is analyzable in terms of utterer's meaning. 

But this is incorrect. Suppose there is a sentence Yof a language L such that Y means 
(pre-theoretically speaking) "Napoleon loves Josephine" (e.g. if L is English, then the 
sentence "Napoleon loves Josephine" will do). When L-speakers wish to mean that 
Napoleon loves Josephine they are more likely to use Y than a sentence Z that means 
(pre-theoretically speaking) "Wisdom is a virtue." To say this is not to say that it is 
impossible for U to mean that Napoleon loves Josephine by uttering Z, it's just to say that 
normally (usually, typically, standardly) U has a much better chance of getting across 
the intended message by uttering Y. Thus it might be suggested that an arbitrary sen- 
tence X means (in L) "Napoleon loves Josephine" iff (roughly) by uttering X, optimally, 
L-speakers mean (would/should mean) that Napoleon loves Josephine. 

Grice is not committed to the absurd position that a hearer must work out what U 
meant by uttering a sentence X in order to work out the meaning of X. To see this as a 
consequence of Grice's theory is to ignore the connection between the theory of con- 
versation and the theory of meaning, ft is Grice's view that typically the hearer must 
establish what U has said (or made as if to say) in order to establish what 17 meant; and 
it is by taking into account the nature and purpose of rational discourse that the hearer 
is able to progress (via, for example, conversational implicature) from what U has said 
(or made as if to say) to what U meant. An analysis of sentence meaning in terms of 
utterer's intentions does not conflict with this idea. 

We must distinguish (1) accounts of what U said and what U meant by uttering X 
and (2) accounts of how hearers recover what U said and what U meant by uttering 
X. What U meant by uttering X is determined solely by U's communicative intentions; 
but of course the formation of genuine communicative intentions by U is constrained 
by U's expectations: U cannot be said to utter X M-intending A to if U thinks that 
there is very little or no hope that U's production of X will result in A 0-ing. If U M- 
intends A actively to entertain the belief that (U thinks) Napoleon loves Josephine, and 
U and A are both English speakers, U may well utter the English sentence "Napoleon 
loves Josephine." To say this is not to commit Grice to the view that sentences that are 
not directly (or so directly) connected to the proposition that Napoleon loves Josephine 
may not be employed to the same effect. 



On the contrary, the theory of conversation is supposed to provide an explanation 
of how this is possible (in the right circumstances). On the assumption that U and A 
are both operating in accordance with the Cooperative Principle and maxims, there 
may well be facts about the context of utterance, the topic of conversation, background 
information, and so on that make it possible for U to mean that Napoleon loves 
Josephine by uttering a very different sentence. U's conception of such things as the 
context of utterance, the topic of conversation, background information, and A's ability 
to work out what U is up to may all play roles in the formation of U's intentions; but 
this does not undermine the view that what determines what U means are U's com- 
municative intentions. 

We can put aside, then, the question of the conceptual coherence of Grice's ana- 
lytical program; the interesting questions concern the adequacy of his concrete 
proposals for explicating sentence meaning and saying. The basic idea is to analyze sen- 
tence meaning in terms of utterer's meaning, and then define saying in terms of a near 
coincidence of utterer's meaning and sentence meaning. Sentence meaning for Grice 
is a species of complete utterance-type meaning, the relevant analysandum for which 
is "X means 'p'," where X is an utterance type and p is a specification of Xs meaning. 
Grice puts forward the following as indicative of the general approach he is inclined to 

IV. For population group G, complete utterance-type X means "p" iff (a) at least some 
(many) members of G have in their behavioral repertoires the procedure of utter- 
ing a token of X if they mean that p, and (b) the retention of this procedure is for 
them conditional on the assumption that at least some (other) members of G have, 
or have had, this procedure in their repertoires. 

For a language containing no context-sensitive expressions, the technical difficulties 
involved in Grice's use of the variable "p" both in and out of quotes can be remedied 
easily enough. But once we turn (as we must) to complete utterance-type meaning for 
a language that contains indexicals such as "I" and "you," demonstratives such as 
"this" and "that," and anaphoric pronouns such as "him" and "her," it is clear that 
some work is needed to transform IV into something acceptable. This is, I think, a very 
serious matter; for without such a transformation, Grice simply will not be able to 
provide an analysis of utterance-type meaning for a language like English, and conse- 
quently he will not be able to provide the sort of definition of saying he wants. 

What, then, is the precise relation between sentence meaning and saying for Grice? 
It might be thought that if we abstract away from the problems raised by indexicals and 
other expressions that highlight the gap between sentence meaning and what is said, 
we will be able to move directly from when uttered by U, X meant "p" to by uttering X, U 
said that p. But there are two problems here. First, only where an utterance-type has 
certain features do we want to say that a token of that type may be used to say some- 
thing. A motorist does not say anything, in Grice's sense, when he indicates an inten- 
tion to make a left turn by signalling. 

Second, certain cases involving, for example, irony or conversational implicature can 
be used to show that we cannot make the relevant move directly. If U utters the sen- 
tence "Smith is an honest man" ironically, although it would be true to say that the 
sentence in question means "Smith is an honest man," it would not be true to say that 



U is saying that Smith is an honest man. On Grice's account, since U does not mean that 
Smith is an honest man (U has no intention of getting A to believe that (he believes 
that) Smith is an honest man) 17 is only making as if to say that Smith is an honest 
man. (Parallel remarks could apply in the case of Professor U's utterance of the 
sentence "Mr X has wonderful handwriting and is always very punctual.") On Grice's 
account, what is said is to be found in the area where sentence meaning and utterer's 
meaning overlap. Abstracting away from context-sensitive expressions once again, 
it looks as though something like the following preliminary definition is on the 
right track: 

V By uttering X, U said that p iff 

(1) by uttering X, part of what U meant was that p 

(2) Xconsists of a sequence of elements (such as words) ordered in a way licensed 
by a system of rules (syntactical rules), and 

(3) X means "p" in virtue of the particular meanings of the elements in X, their 
order and their syntactical structure. 

Grice's unhappiness with V. derives from the existence of conventional implicatures. 
Recall that Grice does not want to allow the sorts of implications that result from the 
use of words such as "but," "yet," "still," "even," and "moreover," to count as part of 
what is said. For example, if U (sincerely and non-ironically) utters the sentence "She 
is poor but she is honest," U does not say that there is some sort of contrast between 
poverty and honesty (or between her poverty and her honesty). Rather, U performs a 
"central speech act," by which U says that she is poor and she is honest, and performs 
in addition a "noncentral speech act," by which U conventionally implicates some sort 
of attitude toward what is said. Putting together what 17 says and what 17 convention- 
ally implicates we get what U conventionally means (see figure 1). So for Grice, at best the 
three conditions in V. define by uttering X, U conventionally meant that p rather than by 
uttering X, U said that p. 

This is as far as Grice goes. He leaves us with the non-trivial task of separating what 
U says and what U conventionally implicates, a rather disappointing terminus. The 
notion of what is said is for Grice a fundamentally important notion in philosophy. If 
this or that philosopher is unclear about what he is saying (as opposed to what he or 
she is implicating) then that philosopher is liable to make all sorts of mistakes, as is 
borne out, Grice thinks, by the crude way in which, for instance the causal theory of 
perception and the theory of descriptions have been written off by philosophers con- 
cerned with the nuances of ordinary language. Furthermore, not until what is said and 
what is conventionally implicated are separated can what is conversationally implicated 
be defined in the manner examined earlier. 

So for Grice, an analysis of saying takes on some urgency, and it is unfortunate that 
he does not get any closer to one than he does in producing V above. However, it may 
well be that Grice has brought us as far as we can go without crossing our own paths. 
Recall that he wants what is said to comprise the truth-conditional content of what is 
conventionally meant by someone making a statement; but he cannot appeal directly 
to truth conditions for fear of undermining one part of his project. There may be no 
simple way out of this. At the same time, only one part of Grice's project is threatened: 
the possibility of providing a definition of saying in terms of utterance-type meaning 



and what is meant. No appeal to truth-conditional content is needed in analyses 
of utterer's meaning or utterance-type meaning, and to that extent Grice has certainly 
illuminated these important notions. In so doing, he has also alerted us to a host of 
important distinctions that philosophers, linguists, cognitive scientists, and literary 
theorists ignore at their peril. 


Works by Grice 

1989: Studies in the Way of Words, Cambridge: Cambridge University Press. 
1991: The Conception of Value, Oxford: Clarendon Press. 

Works by other authors 

Avramides, A. (1989) Meaning and Mind, Cambridge, MA: MIT Press. 

Grandy, R. E. and Warner, R. (eds.) (1986) Philosophical Grounds of Rationality: Intentions, 

Categories, Ends, Oxford: Clarendon Press. 
Martinich, A. P. (1984) Communication andReference, New York: Walter de Gruyter. 
Schiffer, S. (19 72) Meaning, Oxford: Clarendon Press. 
Searle, J. R. (1969) Speech Acts, Cambridge: Cambridge University Press. 
Sperber, D. and Wilson, D. (1986) Relevance, Oxford: Blackwell Publishers. 


Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

G.H. von Wright (1916- ) 


Georg Henrik von Wright was born and educated in Helsinki, Finland, where his 
graduate work was supervised by Eino Kaila, a distinguished Finnish philosopher who 
was connected with the Vienna Circle, and who encouraged von Wright in that direc- 
tion. Because of conditions on the continent, however, von Wright went to Cambridge 
in 1939 to study with C. D. Broad, and there he met Wittgenstein and Moore, who, 
together with Kaila, became the major influences on his philosophical development. He 
returned to Helsinki after a year and in 1941 received a Ph.D. for his thesis "The Logical 
Problem of Induction." After the war, he returned to Cambridge as a lecturer and in 
1948 was appointed to the Chair of Philosophy in Cambridge, which Wittgenstein and, 
before him, Moore had held. In 1951 he resigned to return to Helsinki and resume the 
professorship to which he had been appointed in 1946. In 1961 he was appointed to 
the Academy of Finland, which at the time comprised twelve persons, whose lifetime 
membership in the Academy freed them to do their own work. 

He has continued to teach and lecture around the world, notably as Professor-at- 
Large at Cornell, in giving (among others) the Gifford, Tarner, Woodbridge, Tanner, and 
Leibniz Lectures, and in engaging philosophers from many countries in philosophical 
conversation of the highest quality. He is the subject of a volume in the "Library of 
Living Philosophers" (1989), which contains a bibliography of over 400 papers and 
books (several dozen more have appeared since), essays on his work by thirty-one 
philosophers, and a notably instructive and interesting autobiography. 

Von Wright's philosophical career has been marked by his working through one set 
of issues, then leaving it to focus on another set, and so on, and I have ordered my dis- 
cussion of his contributions to reflect those stages in his career. The order is by no means 
exact, however, because certain themes have persisted throughout his work and 
because he often returns to issues when he has new things to say. 

Induction and probability 

Issues related to induction and probability were the focus of his work from his studies 
for his Ph.D. until the late 1940s. The topic of his doctoral dissertation was the "canons 
of induction," which Bacon and Mill used to ascertain causes and effects, and which 
von Wright replaced with the more precise notions of necessary and sufficient condi- 



tions in order to restate and re-evaluate the classical canons. This was extended, cor- 
rected, and related to probability theory in a number of papers and then in his Treatise 
on Probability and Induction, written in 1948 but not published until 1951. He sums up 
this work as follows: 

I have tried to show how the probabilifying effect of evidence on a given hypothesis is an 
isomorphic reflection in numerical terms of a process of eliminating members from a class 
of hypotheses initially competing with the given one. This eliminative procedure is the 
logical core of Mill's canons. My aim can thus be said to have been a unification of the two 
main branches of inductive logic: induction by elimination in the tradition of Bacon and 
Mill, and induction by confirmation in the tradition founded by the Cambridge logicians 
}. M. Keynes, C. D. Broad, and W. E. Johnson and later continued by Carnap and others. 
(1993: 114) 

His publications from this period also include studies of the paradoxes of confirmation, 
a topic to which he has returned (see 1983b), as he has to induction and probability in 
writing articles on each for the 1959 edition of the Encyclopedia Britannica. 

Philosophical logic 

Von Wright's work on induction and probability was not intended as a contribution to 
mathematical theories like those of von Mises or Kolmogorov, but as a philosophical 
study which used formal methods as a way of understanding and improving the con- 
cepts we employ in evaluating certain kinds of empirical hypotheses. The same is true 
of his work in logic: it aimed not to be a contribution to mathematical logic in the style 
of Godel, Tarski, or Church but to be a philosophy of logic and a philosophical logic (see 
TARSKI, CHURCH, GODEL). He took the task of logic generally to be "to describe and sys- 
tematize the principles used in argumentation, inference, and proof," the aim of his 
own work being philosophical reflection on those principles and the concepts they 
involve. In the course of that work, a new aim emerged: to extend the application of 
logic, as it had developed since Frege, to subject matters that traditionally made no 
explicit use of logical symbols and methods. 

The focus of this work was the concept of logical truth, which he began to investi- 
gate by considering how far the Tractarian notion of logical truth as tautological could 
be extended to quantificational logic. He showed that this could be done for simpler 
quantificational structures, and in so doing invented "distributive normal forms," 
which others appropriated for technical uses (which included showing precisely how 
far the notion of a tautology could be extended). This work led to such related topics as 
conditionals, entailment, negation, and the logical antinomies, on which von Wright 
wrote clarifying and stimulating papers of the highest quality, from which much may 
be learned. (The earlier work is in Logical Studies (1957) and the more recent in 
"Philosophical Logic" (1983b).) 

While working on the quantifiers, he noticed a parallelism between the structure of 
"some," "none," and "all" and that of the modal terms "possible," "impossible," and 
"necessary." Just as the negation of "Some S are P" is equivalent to "No S are P," so the 
negation of "P is possible" is equivalent to "P is impossible," and just as the negation of 



"Some S are not P" is equivalent to 'All S are P," so the negation of "Not-P is possible" 
is equivalent to "P is necessary," and so on. A little later, he noticed the same paral- 
lelism for the deontic modal terms "permissible," "forbidden," and "obligatory": "A is 
not permissible" is equivalent to "A is forbidden"; "Not- A is not permissible" is equiva- 
lent to "A is obligatory," etc. His project was to articulate these modalities in logical 
systems, analogous to propositional and predicate logic, which would have precise rules 
for well-formed formulae and valid inferences, which would enable exact determina- 
tion of what a claim entailed and what it contradicted, and which would permit inves- 
tigation of metalogical issues like consistency and completeness. His ideas on how to 
do this for the strict modalities were written in 1950 (published 1951b), while his 
proposals for the second appeared in his paper, "Deontic Logic," in the first issue of 
Mind for 1951 (reprinted 1957). 

Von Wright showed that further concepts exhibited a similar structure, for example, 
the epistemic modalities "undecided," "falsified," and "verified" (if P is undecided, then 
P is not falsified; if P is falsified, then not-P is verified, etc.), and even time and causal- 
ity can be seen as having a kind of modal structure. Von Wright worked on all these 
concepts, showing how logical principles, concepts, and methods could be applied to 
them, thus extending logical investigations into new areas. (For his summary, see 1993 
(essays VI and VII) and 1989.) This inspired many others to apply logical techniques and 
symbolism to diverse philosophical topics, and while this work is too often technically 
or philosophically uninteresting, some of it has been very significant for both logic and 
philosophy. The latter, significant, work has prompted the view that this is the best - 
even the only - way to do analytic philosophy, but that is a view von Wright emphati- 
cally rejects. 

Ethics, norms, and values 

"Deontic Logic" is von Wright's most famous paper, for it created a new subject, which 
was his own in a special sense. It showed how to symbolize various normative claims 
and precisely determine their logical interrelations, and it opened up new claims and 
relations which are difficult to notice outside a logical system. This "logic of norms" 
remains of great interest to legal philosophers, for instance, who wrestle with such 
questions as whether two laws are mutually consistent or whether, and in what sense, 
a system of laws may be consistent or complete. 

Von Wright's interest in the logic of norms was sustained because of its connection 
with issues about truth. He had since his youth been committed to what he calls a "deep 
conceptual gap separating the world of facts from that of norms and values," and he 
continues to hold that belief in its radical form as the view that normative judgments 
are neither true nor false. This raises the problem of how a logic of norms is even pos- 
sible, for such a logic assumes that there are logical relations between norms and that 
there are disjunctive and conditional norm sentences, and those assumptions appear 
to require that norm sentences have truth-value. 

In dealing with this problem, von Wright appeals to the distinction between sen- 
tences which give a norm, and thus are prescriptive, and those which state that a norm 
has been given, and thus are descriptive, which yields two possibilities for constructing 
a logic of norms without assuming that normative judgments are true or false. The first 



is to have a logic, not of norms as such, but of the sentences which describe given norms 
and thus have truth-values. The second is to have a logic of norms proper but do not 
require that logically complex sentences with logical interrelations must have a truth- 
value, so that "Logic has a wider reach than Truth." Von Wright has changed his mind 
more than once on which of these ways is better (see 1963a, 1968, 1984), and more 
recently has suggested a third possibility, which is that deontic logic is "neither a logic 
of norms nor a logic of norm-propositions but a study of conditions which must be 
satisfied in rational norm-giving activity" (1993: 111; cf. 1996). 

Norm and Action (1963a) and An Essay in Logic and General Theory of Action (1968) 
extend these logical investigations to the actions which norms govern and to the 
changes involved in actions (which I discuss below). But the study of norms is also part 
of moral philosophy, which von Wright was pursuing at the same time, largely because 
of his teaching duties; it resulted in The Varieties of Goodness, his main work in moral 
philosophy. Its central claim is that moral senses of "good," and of "right" or "duty," are 
derivative on non-moral uses of the terms. It is a conceptual inquiry, which aims at 
making "fixed and sharp that which ordinary usage leaves loose and undetermined," 
but since it denies that there is any clear distinction between ethics and metaethics, its 
inquiries into how to improve our moral concepts are also meant to be inquiries into 
what moral point of view we ought to adopt. It assumes that there can be "a philo- 
sophical pursuit deserving the name 'ethics', which shares with a common conception 
of 'meta-ethics' the feature of being a conceptual investigation and with a common 
concept of 'normative ethics' the feature of aiming at directing our lives" (1963b: 6). 
The book is Aristotelean in its insightful classification of the varieties of goodness, in 
its taking seriously the notion of virtue, in its taking the human good to be more basic 
than duty, and in its commitment to there being such a thing as practical rationality, 
which is not inferior to the rationality of the exact sciences. These notions are not rare 
today, but they were audacious in 1963, and it is understandable that von Wright 
regards The Varieties of Goodness as the best argued of all his works and the most fun to 

Philosophy of action 

Von Wright's work in philosophy of action stemmed from his attempt to get a symbol- 
ism adequate to distinguish the various kinds of action governed by norms: to distin- 
guish, for example, what agents do (either by bringing something about or preventing 
something) from what they do not do (either intentionally or by simply doing nothing). 
It struck him that such distinctions could not be expressed without a symbolism for dif- 
ferent kinds of change, and hence he worked at embedding a logic of action in a logic 
of change, a project to which he has returned a number of times (196 3a, 1968, 1973). 
Broader issues about action were implicit in this work and became explicit in 
Explanation and Understanding (1971). It argued that explanations of intentional action 
are logically distinct from explanations in the natural sciences because the latter appeal 
to causal laws while the former invoke conceptual connections between an agent's 
reasons and his actions. Explanation of action is, therefore, necessarily connected with 
practical reason (see 1983a), and the intentional attitudes that function as reasons 
should be seen, not as internal states with causal powers, but as ways of under- 



standing and articulating what agents mean by their behavior. He connected this view 
with the Aristotelean tradition, contrasting it with the Galilean tradition which assim- 
ilated explanation of action (and of historical events) to the law-based model of the 
natural sciences. 

Explanation and Understanding reinforced the rejection of a causal model of action by 
arguing that the concept of cause is inseparable from the concept of (experimental) 
intervention in nature, and hence inseparable from, indeed derivative on, concepts of 
action and agency. This implied that the kind of determinism which threatens to under- 
mine the possibility of genuine human agency is self-defeating, a point von Wright has 
articulated in a number of ways (1973). indeed, determinism has been a core interest 
of his since Explanation and Understanding, which has been expressed partly in logical 
investigations into issues such as kinds of necessity and knowledge and truth about 
the future (1984), partly in discussions of explanation and free will (1973), partly in 
criticisms of the way determinism encourages a "reified conception" of people and 
social institutions as governed by universal laws (1983a, 1993). 

Philosophy of mind 

Von Wright first encountered philosophy as a youth through the mind-body problem, 
but he wrote nothing about it until his Tanner Lectures, "Of Human Freedom" in 1984 
(reprinted in In the Shadow of Descartes, 1998), when reflection on the relation of 
neuro-physiological explanation and action explanation led to some remarks on 
psycho-physical parallelism. He then began to write extensively on the philosophy of 
mind but published little until his 1998 book, which has papers on the relation of phys- 
iological and action explanation, on the concepts of quality and thing, and on percep- 
tion and sensation (with special attention to sounds). The papers do not form a unity 
and do not discuss current literature and controversies (of which von Wright is by no 
means ignorant), but they manifest an intense effort to get clear on some of the most 
difficult and fundamental issues in philosophy of mind, and they contain a wealth of 
distinctions and observations which may prove productive. 


Von Wright knew Wittgenstein well, was named as one of his literary executors, and 
has devoted enormous time and effort to his large and extraordinarily complex 
Nachlass. He has searched for lost material, interviewed persons who knew 
Wittgenstein, organized and indexed the papers (see Wittgenstein, von Wright 1982), 
and edited several volumes of his correspondence and manuscripts. While he has 
written relatively few papers on Wittgenstein's thought, these volumes include some 
splendid studies, which have been published (along with his moving "Biographical 
Sketch") in his 1982 book. 

While von Wright was deeply impressed by his encounters with Wittgenstein, who 
no doubt influenced his work in numerous ways, he has chosen his own path. His work 
on philosophy of action and of mind are the most Wittgensteinian in content (his 
logical work is the least), but none of this can be traced back in any direct way to 
Wittgenstein's writing. All of his work, however, manifests Wittgenstein's stress on con- 



ceptual multiplicity and his distrust of simple answers and philosophical theses, and 
it shares Wittgenstein's aversion to superficial, careless, or pretentious writing (see 



Von Wright also shares with Wittgenstein a sensitivity to the larger context of philos- 
ophy, whose present character he also tends to think of as "the darkness of this time." 
He has written numerous essays (mostly in Swedish or Finnish) on writers such as 
Spengler, Toynbee, Dostoevsky and Tolstoy and on topics including education, the state 
of the humanities, the image of science, the myth of progress, and the idea of revolu- 
tion: essays aimed at clarifying the meaning of his life as a human being rather than 
as a professional philosopher. (See The Tree of Knowledge, 1993.) 

He began as a believer in the "spirit of scientific rationality," which was inspired 
above all by post-Cartesian mathematics and physics, and which in turn inspired the 
philosophy of Kaila and the logical empiricists, the new logic, and analytic philosophy. 
He was never a believer in the idea that scientific rationality would result in inevitable 
progress - even his youthful "aesthetic humanism" was strongly affected by Spengler 

- but he was a believer in the ideal of philosophy as one of the exact sciences. 

This attitude changed when he began teaching moral philosophy, and under the 
influence of Jaeger's Paidea he abandoned "aesthetic humanism" for an individualistic 
"ethical humanism," which recognized a practical rationality on a par with the ratio- 
nality of the exact sciences. This was shattered by the effect the Vietnam war had on 
him, a war he protested in eloquent and effective ways and which forced him to think 
about the human condition in social and political terms. The result was the "social 
humanism" to which he remains committed and which he has expressed in powerful 
essays on the threat to human life of the very scientific rationality which inspired his 
early efforts in philosophy, essays which have made him Scandinavia's most prominent 

- if controversial - public intellectual. 

His critical attitude toward the ideal of scientific rationality, above all for the major 
role it has played in creating social and technological conditions which both threaten 
the natural environment and inspire irrational and unjust ways to try to meet the 
threat, have also changed his conception of philosophy. He thinks of it, not as an exact 
science, but as critical reflection on the underlying assumptions of our thinking, 
judging, and acting -reflection which, since these assumptions are embedded in social 
institutions and traditions, must inevitably scrutinize the foundations of society (see 
"Intellectual Autobiography," 1989). 

Although von Wright thinks of his humanist essays as parallel to rather than part 
of his philosophical work, they exhibit many of the same virtues. They are rich with 
illuminating insights, unexpected observations, and stimulating suggestions for further 
thought, and they are written in a style perfectly suited to their subject and to their 
author. His strictly philosophical work adds such further virtues as exact and intricate 
argument, systematic distinctions, and technical sophistication. These are the virtues 
prized by analytic philosophy, which von Wright has not abandoned, in spite of reject- 
ing scientific rationality as the ideal of philosophy and being skeptical about its role as 
an ideal for any inquiry. What he regards as the chief virtue of philosophy has been 



best expressed in the epigram to his autobiography, which is from Melville's Moby Dick: 
'All deep, earnest thinking is but the intrepid effort of the soul to keep the open 
independence of her sea." 

Bibliography of von Wright's work 

1941: The Logical Problem of Induction, Helsinki: Acta Philosophica Fennica (2nd revised edn., 

Oxford: Blackwell Publishers, 195 7). 
195 la: A Treatise on Induction and Probability, London: Routledge and Kegan Paul. 
1951b: An Essay in Modal Logic, Amsterdam: North Holland. 
1957: Logical Studies, London: Routledge and Kegan Paul. 
1963a: Norm and Action: A Logical Inquiry, London: Routledge and Kegan Paul. 
1963b: The Varieties of Goodness, London: Routledge and Kegan Paul. 

1968: An Essay in Deontic Logic and the General Theory of Action, Amsterdam: North Holland. 
1971: Explanation and Understanding, London: Routledge and Kegan Paul. 
1973: Causality and Determinism, New York: Columbia University Press. 
1981: Freedom and Determination, Helsinki: Acta Philosophica Fennica. 
1982: Wittgenstein, Oxford: Blackwell Publishers. 

1983a: "Practical Reason," Philosophical Papers, vol. 1, Oxford: Blackwell Publishers. 
1983b: "Philosophical Logic," Philosophical Papers, vol. 2, Oxford: Blackwell Publishers. 
1984: "Truth, Knowledge, and Modality," Philosophical Papers, vol. 3, Oxford: Blackwell 

1989: "Intellectual Autobiography," in The Philosophy of Georg Henrik von Wright, The Library of 

Living Philosophers, La Salle, IL: Open Court. 
1993: The Tree of Knowledge and Other Essays, Leiden: Brill. 
1996: Six Essays in Philosophical Logic, Helsinki: Acta Philosophica Fennica. 
1998: In the Shadow of Descartes: Essays in the Philosophy of Mind, Dordrecht: Kluwer Academic 



Blackwell Companions to Philosophy: A Companion to Analytic Philosophy 

Edited by A. P. Martinich, David Sosa 
Copyright © Blackwell Publishers Ltd 2001 

Roderick Chisholm (1916-1999) 

Part I: Epistemology 


Part II: Metaphysics 


Roderick Chisholm's work spans six decades and an impressive range of subjects. His 
books and articles on Brentano and Meinong, together with his work as a translator 
and editor, and as director of the Brentano Foundation, brought Anglo-American ana- 
lytic philosophy back into contact with the riches of the Austrian philosophical tradi- 
tion. He wrote several important papers on the foundations of ethics and axiology (e.g. 
Chisholm 1963, 1974). But Chisholm is best known for his many contributions to 
epistemology and metaphysics. 

Part I: Epistemology 

The most important of Roderick Chisholm's writings on epistemology are Perceiving, 
The Foundations of Knowing, and the three editions of Theory of Knowledge. In these and 
in his other works, Chisholm addressed virtually every major problem in epistemology. 
At the heart of his epistemological system is a set of epistemic principles that are 
intended to generate intuitively plausible results about the degree to which various 
propositions are justified for an individual. The key to Chisholm's epistemology is under- 
standing how these principles fit together and also understanding both their epistemol- 
ogical status (how is it that we can come to know them?) and their metaphysical status 
(are they necessary or contingent, and what is it that makes them true?). 

Terms of epistemic appraisal 

In formulating his epistemological principles, Chisholm presents a set of terms of epis- 
temic appraisal, which he defines using a basic, prephilosophical notion of justification. 
The following are simplified versions of the definitions that appear in the third edition 
of Theory of Knowledge (1989a). 



Certain: A proposition p is certain for an individual S if and only if no other proposi- 
tion is more justified for S to believe. 

Evident: A proposition p is evident for S if and only if S is at least as justified in 
believing p as withholding judgment on that which is counterbalanced. 

Beyond reasonable doubt: A proposition p is beyond reasonable doubt for S if and only 
if S is more justified in believing p than withholding judgment on p. 

Epistemically in the clear: A proposition p is epistemically in the clear for S if and only 
if S is at least as justified in believing p as withholding judgment on p. 

Probable: A proposition p is probable for S if and only if S is more justified in believing 
p than disbelieving p. 

Counterbalanced: A proposition p is counterbalanced for S if and only if S is as 
justified in believing p as believing not-p, and vice-versa. 

Chisholm intends the first five of these terms to be such that the higher ones imply 
the lower ones, and he introduces axioms to ensure this (1989a: 12, 13, 17). So, if a 
proposition is certain for someone, it is also evident for that person, and if it is evident 
for the person, it is also beyond reasonable doubt, and so on down the list. 

Epistemic principles 

Making use of the above terms of epistemic appraisal, Chisholm proposes a set of 
epistemic principles. The principles are expressed as conditionals, whose antecedents 
describe sufficient logical conditions for the application of these terms of epistemic 
appraisal. In the most straightforward case, a principle will assert that if certain non- 
epistemic conditions are satisfied (e.g. conditions about what someone is experiencing, 
believing, etc.), then a proposition p has a certain epistemic status for the person (e.g. 
it is evident or beyond reasonable doubt). 

Chisholm's project in formulating these principles can be compared to a traditional 
project in ethics. A central aim of theoretical ethics is that of describing a set of 
non-moral conditions that is sufficient to make an action morally right. According 
to utilitarians, the non-moral conditions are ones having to do with the production 
of pleasure and the avoidance of pain. If of all the alternatives available to me, 
alternative X will produce the greatest balance of pleasures over pains, then I 
am required to do X. Utilitarians claim that this is the fundamental principle of moral- 
ity. For them, there is but one source of moral obligation. Others disagree, insisting 
that there are other sources as well, ones that are not directly concerned with 
the maximization of happiness. Equality and fairness are among the usual candidates. 
If doing X would produce a fair result, then, according to this view, I have a prima- 
facie obligation to do X even if doing so would not maximize happiness. There 
may be other sources as well, and corresponding to each of these sources will be 
an ethical principle, asserting that the source in question produces a prima-facie 
moral obligation. 

This latter view is the counterpart of Chisholm's view in epistemology. He thinks 
that there is more than one source of epistemic justification, and corresponding to each 
of these sources is an epistemic principle describing the conditions under which 
the source produces justification. However, Chisholm believes that some of these 
sources produce justification only in conjunction with other sources. Thus, the epis- 



temic principles corresponding to these sources must make reference to the workings 
of other principles. The result is a collection of principles that are interdependent in 
complex ways. 

Below are some of the most important of the epistemic principles that Chisholm 

1 If F is a self-presenting property and if S has F and if S believes himself (herself) to 
have F, then it is certain for S that he (she) has F. 

2 If it is evident to S that he (she) is appeared to (|)-ly and it is epistemically in the clear 
for S that something is appearing to him (her) in this way, then it is evident for S 
that something is appearing § to him (her). 

3 If it is evident to S that S is appeared to (|)-ly and if S believes that it is a G that is 
appearing to him (her) in this way and if this proposition is epistemically in the clear 
for S, then it is beyond reasonable doubt for S that he (she) perceives a G. 

4 If S believes a proposition that is not disconfirmed by the set of propositions that 
are evident for S, then the proposition is probable for S. (According to Chisholm, p 
disconfirms q amounts to p tends to make not-q probable.) 

5 If S believes a proposition that is not disconfirmed by that which is probable for S, 
then the proposition is epistemically in the clear for S. 

6 If there are three or more concurrent propositions and if each of them is epistem- 
ically clear for S and if in addition one of them is beyond reasonable doubt for S, 
then they all are beyond reasonable doubt for S. 

7 If there are three or more concurrent propositions and if each of them is beyond 
reasonable for doubt for S and if in addition one is evident for S, then they are all 
evident for S. 

Principle (1) makes reference to "self-presenting properties," which Chisholm takes 
to be purely psychological properties. These properties are non-relational, in the sense 
that from the fact that I have a property of this sort, nothing logically follows about 
how I am related to the non-psychological world. For example, from the fact that I have 
the property of thinking about sailing my boat, it does not follow that I am in fact sailing 
my boat. I does not even follow that I have a boat. Nor, says Chisholm, does anything 
else follow about the non-psychological world. On other hand, from the fact that I have 
the property of being stuck in a traffic jam and thinking about sailing my boat, some- 
thing does follow about the non-psychological world. It follows that there are traffic 
jams, that I am in one, and so on. So, this property is not a self-presenting one. 

Chisholm distinguishes two kinds of self-presenting properties: intentional proper- 
ties (ways of thinking, hoping, fearing, wondering, wishing, desiring, intending, etc.) 
and sensible properties (ways of being appeared to by the various senses). Principle (1) 
says that if I have a self-presenting property and if I believe that I have it, then the 
proposition that I have the property is maximally justified for me. Nothing is more jus- 
tified for me to believe. 

The self-presenting provides a foundation on the basis of which other contingent 
propositions can come to have justification. Principle (2) describes one way of this hap- 
pening. If it is evident to me that I am having a visual experience of the sort that is 
involved in seeing a cat and if in addition it is epistemically in the clear for me that 
something is appearing to me in this way, these two things combine to make it evident 



for me that something is appearing to me in this way. It may not be evident to me 
whether it is a cat or a dog or a bush that is appearing to me, but it is evident for me 
that something is doing so. It is evident, in other words, that I am not hallucinating. 

Chisholm also proposes principles of "perceptual taking." For example, the above 
principle (3) implies that if it is evident to me that I am appeared to in a certain way 
and if I believe that it is a cat that is appearing to me in this way and if moreover this 
proposition is epistemically in the clear for me, then these three things combine to make 
it beyond reasonable doubt for me that I perceive a cat. There is also a principle analo- 
gous to (3) for memory, expressed in terms of what I seem to remember (1989a: 68). 

The antecedents of principles (2) and (3) make reference both to propositions that 
are evident and propositions that are being epistemically in the clear. Principle (1) 
describes how propositions can become evident for me, but Chisholm believes that the 
set of propositions that are epistemically in the clear is much larger than the set of 
evident propositions. Thus, there must be some other source of epistemic justification 
for them. What is this other source? 

Chisholm says that it is belief itself, that one way in which a proposition can obtain 
a degree of epistemic justification is by being believed. Principles (4) and (5) are meant 
to describe how. According to (4), if I believe a proposition that is not disconfirmed by 
the set of propositions that are evident for me, then the proposition is probable for me. 
A large number of propositions can become probable for me in this way. They will have 
this weakly favorable epistemic status even if there is no other positive source of jus- 
tification for them - from self-presentation, perception, or memory, for example. 
Moreover, principle (5) allows these propositions to rise to an even higher epistemic 
status. According to (5), if I believe a proposition that is not disconfirmed by the set of 
other propositions that are probable for me, then this proposition is epistemically in the 
clear for me. What are these propositions that are at least probable for me? In large part, 
they are propositions that satisfy the antecedent of principle (4), namely, believed 
propositions that are not disconfirmed by that which is evident for me. So, (4), as it were, 
creates much of the material for (5) to do its work. 

Principles (4) and (5) are principles of negative coherence. Together they imply that 
if a believed proposition is not incoherent with the set of other propositions that are 
probable for me (many of which get this status by the fact that I believe them and they 
are not disconfirmed by that which is evident for me), then it is acceptable for me to 
believe the proposition. 

With these principles in hand, reconsider the question of how the propositions men- 
tioned in the antecedent of these principles (2) and (3) get the status of being epistemi- 
cally in the clear for me. Principles (4) and (5) provide an answer. They can get this 
status by being believed by me. If I believe a proposition of the sort mentioned in the 
antecedent of (2), say the proposition that something is appearing to me in a cat-like 
way, and if the propositions that are probable for me do not disconfirm this proposition, 
then the proposition is epistemically in the clear for me. And then, principle (2) says 
that this in conjunction with the fact that it is evident to me that I am appeared to in 
a cat-like way makes it evident that something is appearing in a cat-like way to me. It 
is evident that I am not hallucinating. 

Similarly for propositions of the sort mentioned in the antecedent of (3): if I believe 
that it is a real cat appearing to me in a cat-like way and if this proposition is not dis- 



confirmed by the set of other propositions that are probable for me, then the proposi- 
tion is epistemically in the clear for me. And then, principle (3) says that this in con- 
junction with the fact that it is evident for me that I am appeared to in a cat-like way 
makes it beyond reasonable doubt that it is a cat - and not, say, a dog or a bush - that 
I am perceiving. 

Chisholm also thinks that relations of positive coherence among a set of proposi- 
tions, or what he calls "concurrence relations," are an important source of justifica- 
tion. A set of propositions is concurrent just if the propositions are logically 
independent and mutually supportive, in the sense that each proposition in the set is 
such that the others tend to make it probable. 

Chisholm defends two principles of concurrence. Principle (6) says that if there is a 
set of concurrent propositions each of which is epistemically in the clear for me and at 
least one of which is also beyond reasonable for doubt for me, then they all become 
beyond reasonable doubt for me. Principle (7) says something similar for concurrent 
propositions of the next highest epistemic status. According to (7), if there is a con- 
current set of propositions each of which is beyond reasonable for doubt for me and at 
least one of which is evident for me, then they all become evident for me. 

So, despite his reputation as the leading foundationalist, Chisholm is also a coher- 
entist. But unlike a pure coherentist, he does not think that positive coherence relations 
are the only source of empirical justification. 

Together, the above principles describe what Chisholm takes to be some of the prin- 
cipal sources of empirical justification: namely, self-presentation, perception, memory, 
belief coupled with a lack of negative coherence, and, finally, positive coherence among 
propositions with some antecedent positive epistemic status. 

The epistemological and metaphysical status of the principles 

According to Chisholm, we have at least a vague, prephilosophical idea of what it is for 
a belief to be justified (1989a: 5), an idea which guides us in identifying instances of 
beliefs that are justified. In turn, these intuitions about justified beliefs allow the epis- 
temological project to get off the ground. Chisholm is a particularist when it comes to 
matters of epistemological method (1989a: 7). He begins by examining particular 
instances of beliefs that he takes to be justified, and he then tries to abstract out of these 
instances general conditions of justification, which he expresses in the form of epis- 
temic principles. 

Chisholm also presupposes that we can improve and correct our beliefs by reflection, 
eliminating those that are unjustified and adding others that are justified (1989a: 1, 
5). This presupposition acts as a constraint when he tries to use particular instances of 
justified belief to formulate general conditions of justification. It forces him to look for 
conditions to which we have reflective access, since otherwise there would be no reason 
to think that we could eliminate unjustified beliefs and add justified ones simply by being 
reflective. This is one of the senses in which Chisholm is an internalist about justifica- 
tion, in an epistemic sense. 

The prephilosophical notion of justification that allows epistemology to get off the 
ground is vague, like most ordinary notions, but it need not remain so. One of the ben- 
eficial by-products of formulating and refining epistemic principles is that the basic 



notion becomes increasingly precise, so that eventually epistemolgists are in a position 
to give a general characterization of it. According to Chisholm, the characterization is 
to be given in ethical terms. Epistemic justification is ultimately to be understood in 
terms of ethical requirements on our believings and withholdings. More specifically, to 
say that an individual S is more justified in believing p than withholding on p is to say 
that S is required to prefer the former over the latter (1989a: 59). Chisholm goes on to 
claim that requirements to prefer are best explicated in a negative way. The requirement 
to prefer believing p over withholding on p is a requirement not to choose between 
believing and withholding without choosing the former, and this, he points out, is a 
requirement that can be satisfied even if one does not have direct control over one's 
believings and withholdings. 

In addition, Chisholm says that this requirement is one that supervenes on non- 
normative states, specifically, on conscious states (1989a: 60). As such, a proposi- 
tion could not have an epistemic status different from the one it does have for an indi- 
vidual without that individual's psychological states being different. Thus Chisholm 
takes his epistemic principles to express necessary truths, and the truths that 
they express are ultimately ones about the relationship between an individual's con- 
scious psychological states at a time and an ethical requirement on believings and 

This illustrates another sense in which Chisholm is an internalist about justification, 
a metaphysical sense. The conditions that make a proposition evident or beyond rea- 
sonable doubt or probable are internal conditions. They are current, psychological 
states, not non-psychological "external" states, and not past psychological states. 
Chisholm's epistemic internalism requires something in addition to this. It requires that 
we always have reflective access to these internal conditions. 

The definition of knowledge 

The epistemic principles and the terms of epistemic appraisal used to formulate the 
principles constitute the heart of Chisholm's epistemological system. They are the tools 
Chisholm uses to address the major questions of epistemology. Among these questions, 
none has preoccupied Chisholm more than the question, What is knowledge? 

Over his career, he proposed various definitions of knowledge, most of them vari- 
ants of the idea that knowledge is non-defectively evident true belief. Like many pro- 
posed definitions of knowledge, Chisholm's definition was aimed at coming to grips with 
a pair of examples presented by Edmund Gettier, which were designed to illustrate that 
knowledge cannot be adequately defined as justified true belief. The basic idea behind 
both counterexamples is that one could be justified in believing a falsehood P, from 
which one deduces a truth 0. In this case one has a justified true belief in but does 
not know 0. Gettier's examples inspired a host of similar counterexamples, and the 
search was on for a fourth condition of knowledge, one that could be added to belief, 
truth, and justification to produce an adequate analysis of knowledge. 

The two most distinctive aspects of Chisholm's attempt to handle Gettier problems 
are, first, his insistence that a belief must be evident to count as knowledge, and second, 
his insistence that what makes the belief evident must be non-defective (Chisholm adds 
some further qualification; see 1989a: 98). 



"Evident" is among the strongest of Chisholm's terms of epistemic appraisal, 
ranking only below that which is certain. A proposition is certain for an individual 
S only if it is maximally justified; no other proposition is more justified for S to believe. 
Among the propositions that can be certain are simple necessary truths, for example, 
the elementary truths of arithmetic, as well as contingent propositions about self- 
presenting states. In defining knowledge in terms of the evident, Chisholm is rejecting 
the view that knowledge requires certainty. On the other hand, he is insisting that 
knowledge involves a very high degree of justification. For Chisholm, a paradigmatic 
requirement on believing is that of withholding judgment on that which is counter- 
balanced, for example, the proposition that the next toss of a fair coin will turn up 
heads. A proposition is evident for S, in turn, only if S is at least as justified in believing 
it as withholding judgment on that which is counterbalanced. For Chisholm, this rep- 
resents a very high degree of justification. 

Moreover, if S is to have knowledge of a proposition, not only does the proposition 
have to be evident for S, in addition that which provides this very high degree of justi- 
fication must be non-defective, in the sense that it must not make any falsehood evident 
for S. To illustrate the intuitive force of this requirement, consider one of Gettier's exam- 
ples. Smith has very strong evidence for the proposition that Jones owns a Ford, since 
Smith is aware that Jones has always owned a car, that the car has always been a Ford, 
that Jones has just offered Smith a ride while driving a Ford, and so on. From this evi- 
dence and the proposition that Jones owns a Ford, Smith deduces the disjunctive propo- 
sition, either Jones owns a Ford or Brown is in Barcelona. He infers this proposition 
despite having no idea of where Brown is. However, it turns out that Jones does not in 
fact own a Ford (he has been driving someone else's car) while, by chance, Brown is in 
Barcelona. So Smith has very strong justification for the proposition that either Jones 
owns a Ford or Brown is in Barcelona; he believes the proposition; and the proposition 
is true. But it seems as if Smith does not know the proposition. Why? Chisholm's answer 
is that although the proposition may be both true and evident for Smith, the consider- 
ations which makes the proposition evident for him also makes evident a falsehood, 
namely, that Jones owns a Ford. 

Part II: Metaphysics 

Whereas Chisholm's epistemological views constitute a unified whole that may be use- 
fully and concisely summarized, the many metaphysical problems he addressed form 
a heterogeneous collection that does not submit readily to concise overview. 
Furthermore, his role in the rejuvenation of metaphysics during the second half of the 
twentieth century would not be conveyed by a summary of his approaches to particu- 
lar problems. One must back up a bit to see why and where his influence was great. 

Chisholm's impact on contemporary metaphysics would be hard to overestimate. By 
the end of 1950s his contributions were already numerous. He mounted an influential 
defense of the meaningfulness of traditional metaphysical questions in the face of defla- 
tionary critiques from the "ordinary language" philosophers (1951, 1952, 1964). He 
helped bring down the curtain on phenomenalism (1948, 1957b: Appendix). He drew 
attention to Franz Brentano's characterization of the psychological in terms of "inten- 
tional inexistence," and attempted to rehabilitate it as a logico-linguistic criterion of 



sentences reporting intentional mental states (195 5-6). He defended the thesis that lin- 
guistic intentionality is to be explicated in terms of the intentionality of thought, and 
not the reverse (1955a, 1957a). And he helped focus debates about counterfactuals, 
dispositions, and laws of nature (1946, 1955b). This work was widely anthologized in 
subsequent decades. 

But his most important contributions to metaphysics came somewhat later, as he 
began to construct complex, evolving, and interconnected theories of action, persis- 
tence through time, events and causation, reference and intentionality, and ontologi- 
cal categories. 

A summary of the positions he defended on these topics would go some way toward 
explaining his importance as a metaphysician. In many areas, one still finds Chisholm's 
work cited as containing the paradigmatic formulation of an important position, or the 
original statement of a paradox that stands in need of resolution. In action theory, for 
instance, his defense of the incompatibility of freedom and determinism, and of agent 
causation, are as frequently discussed as ever. But a simple survey of Chisholm's views 
on particular metaphysical issues would miss the forest for the trees. A better picture 
of his place in twentieth-century metaphysics can be gained by considering the status 
of metaphysics at the time his career began, and by comparing Chisholm's methodol- 
ogy with that of another imposing figure from the same generation: W. V Ouine. 

Metaphysics at mid-century 

Most philosophical work that bears the (sometimes pejorative) label "metaphysics" is 
characterized by its attention to matters of ontology. A central part of the discipline has 
always been the construction of comprehensive ontological schemes, theories about 
the nature of and relations among the most abstract categories under which absolutely 
everything falls, together with the explicit use of these ontological distinctions in the for- 
mulation of solutions to philosophical problems. Indeed, one could argue that distinc- 
tively metaphysical problems always involve the very abstract categories appropriate to 
ontology; and that any philosophical problem becomes, at least in part, a metaphysical 
problem as soon as ontological distinctions become central to its statement and 

By mid-century, metaphysics of this sort had fallen on hard times. The air had gone 
out of debates about the ontological status of universals and particulars, the distinc- 
tion between essence and accident, and so on. Russell and Moore and, perhaps, the trac- 
tarian Wittgenstein had taken such questions pretty seriously (see MOORE, RUSSELL, and 
WITTGENSTEIN); but logical empiricism, Wittgensteinian "therapy" and Austinian "ordi- 
nary language philosophy" had eclipsed the metaphysical preoccupations of the earli- 
est "analytic philosophers" (see AUSTIN and WITTGENSTEIN). The proponents of these 
influential doctrines all thought (albeit for different reasons) that the traditional ques- 
tions of metaphysics were misguided, unanswerable, nonsensical. 

Further, the reputation of metaphysics was poorly served by the obscurity of many 
of its most well-known practitioners. Clarity of exposition was not among the virtues 
exemplified by Royce, Bradley, Bosanquet, Bergson, Whitehead - names that meant 
metaphysics at the time. To the skeptical, it could easily seem that the recipe for success 
in metaphysics was this: (1) invent your own baroque ontological scheme, using a new, 



peculiar jargon; (2) claim that it is radically opposed to all preceding metaphysical 
systems; and (3) explain its intricacies by the introduction of further undefined tech- 
nical terms in a series of ever longer books. Chisholm's chief contribution to contem- 
porary metaphysics was to show, by precept and, more importantly, by example that it 
is possible to construct metaphysical systems on a grand scale without falling into these 
vices. He championed a chastened approach to metaphysics, one that neither shies 
away from the traditional problems of ontology, nor falls back into the arcane, unteth- 
ered system-building that had given metaphysics a bad name. 

The comparison with Ouine 

W. V Ouine began teaching at Harvard while Chisholm was a graduate student. Ouine 
provided something that would prove crucial to Chisholm's metaphysical program: the 
approach to questions of ontological commitment defended in "On What There Is" 
(1948), but already in place by 1939, when Chisholm was a student (Ouine 1939). 
Chisholm took Quine's criterion of ontological commitment to amount to the follow- 
ing injunction: If one affirms a statement using a name or other singular term, or an 
initial phrase of "existential quantification," like "There are some so-and-so's" (see 
OUINE), then one must either (1) admit that one is committed to the existence of things 
answering to the singular term or satisfying the description, or (2) provide a "para- 
phrase" of the statement that eschews singular terms and quantification over so-and- 
so's. Both Ouine and Chisholm agree that Meinong, who affirms truths about all sorts 
of things which he then admits do not exist, is trying to have his cake and eat it too; 
Meinong must be resisted if metaphysics is to be kept honest. 

Chisholm's metaphysics looks nothing like Quine's, however. For Ouine, it is the 
deliverances of science alone that need be taken into account when attempting to 
work out one's ontological commitments; he identifies the project of "limning the true 
and ultimate structure of reality" with that of working out the most ontologically 
austere regimentation of the language of the harder sciences. This enables Ouine to 
keep his ontology lean, including nothing but the most well-understood, sharply 
demarcated things: ultimately, nothing but concrete spatiotemporal entities and the 
abstract but well-defined world of set theory. But the cost is great: the repudiation 
of quite a lot of what we would ordinarily regard as truisms about beliefs, desires, and 
other intentional attitudes; about what must or might be the case; about what would 
have happened if . . . ; and so on (see, e.g., Ouine 1960). Chisholm, however, asks: Why 
not assume, in the seminar room, the same things we take ourselves to know in every- 
day life? Why, when we do philosophy, should we appeal to nothing but what we 
find in our physics and chemistry textbooks? Chisholm rejects Quine's skepticism 
toward all but science; an ontological scheme must show its adequacy on a much 
broader playing field. 

Both Chisholm and Ouine agree that ontological schemes are to be judged by the 
competing desiderata of simplicity and sufficiency of scope. One scheme is simpler than 
another if it posits fewer, and better understood, types of entities. One scheme is supe- 
rior to another in scope insofar as it allows for the statement of satisfactory philosophi- 
cal theories on more subjects, theories that preserve, sometimes in the face of apparent 
contradiction or philosophical puzzlement, most of what we take ourselves to know. 



Quine's austere ontological naturalism is purchased at the cost of severe restrictions on 
the scope of what we may reasonably take ourselves to know. Although one cannot 
accept the mathematics needed for science without set theory no further "queer enti- 
ties" need be recognized by one who affirms nothing but the deliverances of the 
(sufficiently hard) sciences. Chisholm, however, has many more truths to consider; 
for him, balancing the competing desiderata of simplicity of scheme and sufficiency of 
scope is much trickier. The adequacy of an ontological scheme comes to turn upon 
its role in the resolution of the traditional problems of philosophy, most of which Quine 
was able to sidestep by rejecting the commonsensical convictions from which the 
problems arise. 

It is no surprise, then, to find the two philosophers differing drastically despite their 
initial point of agreement. Chisholm finds that one cannot arrive at metaphysical the- 
ories satisfying both desiderata of simplicity and scope without making reference to 
things not found in Quine's ontology, such as "intensional objects." He can find no onto- 
logically perspicuous theory that does justice to what we know about persons while 
eschewing irreducibly intentional (psychological) notions (e.g. "conceiving," "attribut- 
ing"). Ultimately, he concludes that persons must be very special indeed: they have 
causal powers unlike those found elsewhere in nature, they can "grasp" or conceive of 
abstract objects, and their persistence conditions are mysteriously different from those 
of ordinary physical objects. Quine, and many other naturalistic ally inclined philoso- 
phers, will find such conclusions fantastical. Be that as it may, the theories Chisholm 
constructs offer solutions to a host of philosophical problems; and his metaphysical 
program stands as a challenge to be met by those who would be more naturalistic or 
nominalistic than Chisholm, but who are not prepared to retreat into a skepticism as 
radical as Ouine's. 

Chisholmian methodology illustrated: states of affairs as necessary things 

Chisholm's ontological views underwent frequent revision, as one or another scheme 
proved inadequate in scope, unable to make room for enough of what we take ourselves 
to know; or as he thought of some way to keep a plausible philosophical theory in place 
while simplifying its ontological commitments. One of the more radical changes was 
the rejection of "states of affairs" in the early 1980s. It provides a good example of 
Chisholm's effort to make systematic metaphysics responsible by tying ontology to the 
resolution of a wide spectrum of philosophical problems. In this case, the change was 
brought about by problems of self-reference. 

The greatest ontological divide in Chisholm's theory of categories is between 
necessary things and contingent things. The states of affairs so central to Chisholm's 
ontology throughout the 1960s and 1970s were taken to be necessary things. 

Chisholm advanced several ways of marking the distinction between necessary 
and contingent things. He hoped to restrict his modal primitives to those expressible 
by means of one locution: "x is necessarily such that it is F," where "F" can be replaced 
by any predicate, and the phrase is equivalent to "x is necessarily such that it exists 
if and only if it is F." (Something is possibly F, of course, if and only if it is not neces- 
sarily not F.) But then, even if "exists" were allowed as a predicate substitutable for 
F, replacing F with "exists" yields only a sense of "necessarily existing" according to 



which everything exists necessarily. One proposal for making the distinction within 
his restricted vocabulary is this: contingent things are possibly such as to be coming 
into existence or passing away (i.e. possibly such as to have had no properties and 
possibly such as to be going to have no properties) and necessary things are neither 
(cf. 1989b: 164, and 1996: 127). This presupposes that there are no things that could 
have failed to exist but that, given that they do exist, cannot possibly be created or 
destroyed. Some might have doubts about this assumption. Chisholm may have doubted 
it himself, since he tried other ways of making the distinction. 

Another proposed mark of the necessary/contingent divide is this: x is nec- 
essary if and only if x has a property that is essential to it and that nothing else 
could possibly have (i.e. a property that is an individual essence of x); and every- 
thing is such that something has that property (1986: 26). This presupposes that 
every necessary thing has an individual essence. Perhaps Chisholm had doubts 
about this assumption, too; for his last attempt to formulate a criterion for the 
necessary existence of a thing, x, was: "There is an attribute that is such that (1) 
everything is necessarily such that there is something having that attribute, (2) 
x is necessarily such that it has that attribute, and (3) that attribute is not 
necessarily had by everything" (1996: 17). Counterexamples are generated if 
one allows for disjunctive properties such as being either an animal or prime. But 
Chisholm also developed theories of the structure of properties, including accounts 
of what it is for a property to be a disjunction of two others. Perhaps he would 
have found the resources there to refine his last definition so as to rule out such 

Chisholm long held that there were at least two sorts of necessary thing: states 
of affairs and properties or attributes. (He always at least left an opening in his table 
of the categories for a third, as well: God, a necessary substance upon which all 
else depends.) He advocates an "intentional approach" to both states of affairs and 
properties; that is, he claims that their criteria of identity and structural features can 
only be adequately described using intentional terms, such as "believing" and "con- 
ceiving." States of affairs are defined as those things which one may believe (1976: 
117), properties as those things which one may believe to be exemplified by other things 
(1996: 29). Both are given intentional criteria of identity. A state of affairs p is identi- 
cal with a state of affairs q if and only if, necessarily, (1) p "obtains" or "occurs" if and 
only if qdoes (1976: 118); and (2) whoever believes p believes q, and vice versa. A prop- 
erty F is identical with a property G if and only if, necessarily, (1) something exempli- 
fies F if and only if it exemplifies G, and (2) whoever conceives F conceives G, and vice 
versa (1989b: 145). Propositions are identified with states of affairs that either always 
obtain or never obtain, events with obtaining states of affairs that are not propositions 
and that entail the exemplification of a certain sort of property - a property "rooted 
in" the time at which it is exemplified. 

Chisholm's ontology is "realist" in several senses of the term. It includes properties 
that, like Plato's universals, exist whether or not they are exemplified. It is opposed to 
psychologism and linguisticism about the subject matter of logic. Logic discovers nec- 
essary relations among propositions: necessarily existing states of affairs, in no sense 
mind-dependent or language-dependent. Indeed, the true propositions are not to be dis- 
tinguished from facts. And so Chisholm advocates what is sometimes called an "iden- 



tity theory of truth": true propositions "correspond with facts in the fullest sense that 
is possible, for they are facts" (1977: 88). 

The first person and the rejection of haecceities 

Throughout the 1960s and 1970s states of affairs figure prominently in Chisholm's 
metaphysics, epistemology and metaethics. Here are some examples of the many duties 
they perform. Statements about particular occurrences of a given type of event are to 
be paraphrased in terms of the "obtaining" of abstract, eternal states of affairs. And a 
causal relation between a pair of events is really a matter of two states of affairs being 
causally related relative to a certain time. On this approach, there is no need to recognize 
an ontological category of "tropes" or "particularized properties" in addition to states 
of affairs and properties conceived as Platonic universals (1976: ch. iv). "Times" are 
given a gloss much like A. N. Prior's: they are maximal, consistent states of affairs, com- 
plete ways the world could be "all at once" (1979a: 357). Belief and other propositional 
attitudes are said to be relations between thinkers and states of affairs (1976: ch. iv). 
More generally, a relatively simple ontology of properties, states of affairs, and contin- 
gent, persisting individual things appears to be adequate to the formulation of philo- 
sophical theories across the whole range of subjects Chisholm addressed in this period. 

The phenomenon of first person reference subjected the ontology of states of affairs 
to considerable strain. If propositional attitudes are relations between thinkers and 
states of affairs, what states of affairs are implicated in those attitudes expressed using 
the first person pronoun? Ernst Mach catches sight of himself in a mirror without real- 
izing who it is, and thinks: "That is a shabby pedagogue," without thinking: "I am a 
shabby pedagogue." The contents of the two attitudes differ; but how is this difference 
to be reflected as a difference in the structures of states of affairs, as it must be on 
Chisholm's theory? Since states of affairs are necessary things, their constituents, too, 
must be necessary existents. The only way, then, for a state of affairs to be about some 
contingent thing is for it to contain an "individual concept" of that thing: a property 
only one thing could have, and one that is had by that thing. But what individual 
concept is involved in my first person thoughts? What extra property is there in the state 
of affairs I am a shabby pedagogue that is not present in Someone is a shabby pedagogue? 
Surely I need not know anything about my relations to other things in order to think 
a first person thought; so it must be some intrinsic property, peculiar to me, that enables 
me to think of myself in this way. And so Chisholm is led to accept the notion that each 
person has an "haecceity" an individual essence peculiar to him or her, and "repug- 
nant to" everything else (1976: ch. I). 

Chisholm gradually came to feel that introducing haecceities for this purpose was a 
cheat. Although extraordinarily useful, haecceities remain, at bottom, utterly mysteri- 
ous. We cannot rest content with simply positing their existence on the basis of their 
usefulness, since a part of their use is supposed to be their accessibility to intellectual 
grasp by the thinking things that exemplify them: 

If this essentialistic theory were true, then every time a person expresses himself by means 
of an I-sentence he grasps his own essence or haecceity. But, one wonders, do I ever thus 
grasp my own individual essence or haecceity? If I do ever grasp it, shouldn't I be able to 



single out its various marks? . . . [I]f I can grasp my individual essence, then I ought also 
to be able to single out in it those features that are unique to it. If being me is my indi- 
vidual essence and being you is yours, then, presumably, each analyses into personhood 
and something else as well - one something else in my case and another in yours. But I 
haven't the faintest idea what this something else might be. . . . 

I think that Brentano was right a